Commit | Line | Data |
---|---|---|
b8aa1682 JD |
1 | /* |
2 | * Copyright (C) 2012 - Julien Desfossez <jdesfossez@efficios.com> | |
3 | * David Goulet <dgoulet@efficios.com> | |
cd60b05a | 4 | * 2013 - Jérémie Galarneau <jeremie.galarneau@efficios.com> |
5a693d30 | 5 | * 2015 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com> |
b8aa1682 JD |
6 | * |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License, version 2 only, | |
9 | * as published by the Free Software Foundation. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, but WITHOUT | |
12 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
13 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
14 | * more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License along | |
17 | * with this program; if not, write to the Free Software Foundation, Inc., | |
18 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. | |
19 | */ | |
20 | ||
21 | #define _GNU_SOURCE | |
6c1c0768 | 22 | #define _LGPL_SOURCE |
b8aa1682 JD |
23 | #include <getopt.h> |
24 | #include <grp.h> | |
25 | #include <limits.h> | |
26 | #include <pthread.h> | |
27 | #include <signal.h> | |
28 | #include <stdio.h> | |
29 | #include <stdlib.h> | |
30 | #include <string.h> | |
31 | #include <sys/mman.h> | |
32 | #include <sys/mount.h> | |
33 | #include <sys/resource.h> | |
34 | #include <sys/socket.h> | |
35 | #include <sys/stat.h> | |
36 | #include <sys/types.h> | |
37 | #include <sys/wait.h> | |
173af62f | 38 | #include <inttypes.h> |
b8aa1682 JD |
39 | #include <urcu/futex.h> |
40 | #include <urcu/uatomic.h> | |
41 | #include <unistd.h> | |
42 | #include <fcntl.h> | |
43 | #include <config.h> | |
44 | ||
45 | #include <lttng/lttng.h> | |
46 | #include <common/common.h> | |
47 | #include <common/compat/poll.h> | |
48 | #include <common/compat/socket.h> | |
f263b7fd | 49 | #include <common/compat/endian.h> |
e8fa9fb0 | 50 | #include <common/compat/getenv.h> |
b8aa1682 | 51 | #include <common/defaults.h> |
3fd27398 | 52 | #include <common/daemonize.h> |
b8aa1682 JD |
53 | #include <common/futex.h> |
54 | #include <common/sessiond-comm/sessiond-comm.h> | |
55 | #include <common/sessiond-comm/inet.h> | |
b8aa1682 JD |
56 | #include <common/sessiond-comm/relayd.h> |
57 | #include <common/uri.h> | |
a02de639 | 58 | #include <common/utils.h> |
cd60b05a | 59 | #include <common/config/config.h> |
5a693d30 | 60 | #include <urcu/rculist.h> |
b8aa1682 | 61 | |
0f907de1 | 62 | #include "cmd.h" |
d3e2ba59 | 63 | #include "ctf-trace.h" |
1c20f0e2 | 64 | #include "index.h" |
0f907de1 | 65 | #include "utils.h" |
b8aa1682 | 66 | #include "lttng-relayd.h" |
d3e2ba59 | 67 | #include "live.h" |
55706a7d | 68 | #include "health-relayd.h" |
9b5e0863 | 69 | #include "testpoint.h" |
2f8f53af | 70 | #include "viewer-stream.h" |
2a174661 DG |
71 | #include "session.h" |
72 | #include "stream.h" | |
58eb9381 | 73 | #include "connection.h" |
8921c6e4 | 74 | #include "tracefile-array.h" |
b8aa1682 JD |
75 | |
76 | /* command line options */ | |
0f907de1 | 77 | char *opt_output_path; |
b5218ffb | 78 | static int opt_daemon, opt_background; |
3fd27398 MD |
79 | |
80 | /* | |
81 | * We need to wait for listener and live listener threads, as well as | |
82 | * health check thread, before being ready to signal readiness. | |
83 | */ | |
84 | #define NR_LTTNG_RELAY_READY 3 | |
85 | static int lttng_relay_ready = NR_LTTNG_RELAY_READY; | |
86 | static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */ | |
87 | static pid_t child_ppid; /* Internal parent PID use with daemonize. */ | |
88 | ||
095a4ae5 MD |
89 | static struct lttng_uri *control_uri; |
90 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 91 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
92 | |
93 | const char *progname; | |
b8aa1682 | 94 | |
65931c8b | 95 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
96 | static int tracing_group_name_override; |
97 | ||
98 | const char * const config_section_name = "relayd"; | |
65931c8b | 99 | |
b8aa1682 JD |
100 | /* |
101 | * Quit pipe for all threads. This permits a single cancellation point | |
102 | * for all threads when receiving an event on the pipe. | |
103 | */ | |
0b242f62 | 104 | int thread_quit_pipe[2] = { -1, -1 }; |
b8aa1682 JD |
105 | |
106 | /* | |
107 | * This pipe is used to inform the worker thread that a command is queued and | |
108 | * ready to be processed. | |
109 | */ | |
58eb9381 | 110 | static int relay_conn_pipe[2] = { -1, -1 }; |
b8aa1682 | 111 | |
26c9d55e | 112 | /* Shared between threads */ |
b8aa1682 JD |
113 | static int dispatch_thread_exit; |
114 | ||
115 | static pthread_t listener_thread; | |
116 | static pthread_t dispatcher_thread; | |
117 | static pthread_t worker_thread; | |
65931c8b | 118 | static pthread_t health_thread; |
b8aa1682 | 119 | |
5a693d30 MD |
120 | /* |
121 | * last_relay_stream_id_lock protects last_relay_stream_id increment | |
122 | * atomicity on 32-bit architectures. | |
123 | */ | |
124 | static pthread_mutex_t last_relay_stream_id_lock = PTHREAD_MUTEX_INITIALIZER; | |
095a4ae5 | 125 | static uint64_t last_relay_stream_id; |
b8aa1682 JD |
126 | |
127 | /* | |
128 | * Relay command queue. | |
129 | * | |
130 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
131 | * queue. | |
132 | */ | |
58eb9381 | 133 | static struct relay_conn_queue relay_conn_queue; |
b8aa1682 JD |
134 | |
135 | /* buffer allocated at startup, used to store the trace data */ | |
095a4ae5 MD |
136 | static char *data_buffer; |
137 | static unsigned int data_buffer_size; | |
b8aa1682 | 138 | |
d3e2ba59 JD |
139 | /* Global relay stream hash table. */ |
140 | struct lttng_ht *relay_streams_ht; | |
141 | ||
92c6ca54 DG |
142 | /* Global relay viewer stream hash table. */ |
143 | struct lttng_ht *viewer_streams_ht; | |
144 | ||
5a693d30 MD |
145 | /* Global relay sessions hash table. */ |
146 | struct lttng_ht *sessions_ht; | |
0a6518b0 | 147 | |
55706a7d | 148 | /* Relayd health monitoring */ |
eea7556c | 149 | struct health_app *health_relayd; |
55706a7d | 150 | |
cd60b05a JG |
151 | static struct option long_options[] = { |
152 | { "control-port", 1, 0, 'C', }, | |
153 | { "data-port", 1, 0, 'D', }, | |
8d5c808e | 154 | { "live-port", 1, 0, 'L', }, |
cd60b05a | 155 | { "daemonize", 0, 0, 'd', }, |
b5218ffb | 156 | { "background", 0, 0, 'b', }, |
cd60b05a JG |
157 | { "group", 1, 0, 'g', }, |
158 | { "help", 0, 0, 'h', }, | |
159 | { "output", 1, 0, 'o', }, | |
160 | { "verbose", 0, 0, 'v', }, | |
161 | { "config", 1, 0, 'f' }, | |
162 | { NULL, 0, 0, 0, }, | |
163 | }; | |
164 | ||
165 | static const char *config_ignore_options[] = { "help", "config" }; | |
166 | ||
b8aa1682 JD |
167 | /* |
168 | * usage function on stderr | |
169 | */ | |
5a693d30 | 170 | static void usage(void) |
b8aa1682 JD |
171 | { |
172 | fprintf(stderr, "Usage: %s OPTIONS\n\nOptions:\n", progname); | |
994fa64f DG |
173 | fprintf(stderr, " -h, --help Display this usage.\n"); |
174 | fprintf(stderr, " -d, --daemonize Start as a daemon.\n"); | |
b5218ffb | 175 | fprintf(stderr, " -b, --background Start as a daemon, keeping console open.\n"); |
994fa64f DG |
176 | fprintf(stderr, " -C, --control-port URL Control port listening.\n"); |
177 | fprintf(stderr, " -D, --data-port URL Data port listening.\n"); | |
8d5c808e | 178 | fprintf(stderr, " -L, --live-port URL Live view port listening.\n"); |
994fa64f DG |
179 | fprintf(stderr, " -o, --output PATH Output path for traces. Must use an absolute path.\n"); |
180 | fprintf(stderr, " -v, --verbose Verbose mode. Activate DBG() macro.\n"); | |
65931c8b | 181 | fprintf(stderr, " -g, --group NAME Specify the tracing group name. (default: tracing)\n"); |
cd60b05a | 182 | fprintf(stderr, " -f --config Load daemon configuration file\n"); |
b8aa1682 JD |
183 | } |
184 | ||
cd60b05a JG |
185 | /* |
186 | * Take an option from the getopt output and set it in the right variable to be | |
187 | * used later. | |
188 | * | |
189 | * Return 0 on success else a negative value. | |
190 | */ | |
5a693d30 | 191 | static int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 192 | { |
cd60b05a JG |
193 | int ret; |
194 | ||
195 | switch (opt) { | |
196 | case 0: | |
197 | fprintf(stderr, "option %s", optname); | |
198 | if (arg) { | |
199 | fprintf(stderr, " with arg %s\n", arg); | |
200 | } | |
201 | break; | |
202 | case 'C': | |
e8fa9fb0 MD |
203 | if (lttng_is_setuid_setgid()) { |
204 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
205 | "-C, --control-port"); | |
206 | } else { | |
207 | ret = uri_parse(arg, &control_uri); | |
208 | if (ret < 0) { | |
209 | ERR("Invalid control URI specified"); | |
210 | goto end; | |
211 | } | |
212 | if (control_uri->port == 0) { | |
213 | control_uri->port = DEFAULT_NETWORK_CONTROL_PORT; | |
214 | } | |
cd60b05a JG |
215 | } |
216 | break; | |
217 | case 'D': | |
e8fa9fb0 MD |
218 | if (lttng_is_setuid_setgid()) { |
219 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
220 | "-D, -data-port"); | |
221 | } else { | |
222 | ret = uri_parse(arg, &data_uri); | |
223 | if (ret < 0) { | |
224 | ERR("Invalid data URI specified"); | |
225 | goto end; | |
226 | } | |
227 | if (data_uri->port == 0) { | |
228 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
229 | } | |
cd60b05a JG |
230 | } |
231 | break; | |
8d5c808e | 232 | case 'L': |
e8fa9fb0 MD |
233 | if (lttng_is_setuid_setgid()) { |
234 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
235 | "-L, -live-port"); | |
236 | } else { | |
237 | ret = uri_parse(arg, &live_uri); | |
238 | if (ret < 0) { | |
239 | ERR("Invalid live URI specified"); | |
240 | goto end; | |
241 | } | |
242 | if (live_uri->port == 0) { | |
243 | live_uri->port = DEFAULT_NETWORK_VIEWER_PORT; | |
244 | } | |
8d5c808e AM |
245 | } |
246 | break; | |
cd60b05a JG |
247 | case 'd': |
248 | opt_daemon = 1; | |
249 | break; | |
b5218ffb MD |
250 | case 'b': |
251 | opt_background = 1; | |
252 | break; | |
cd60b05a | 253 | case 'g': |
e8fa9fb0 MD |
254 | if (lttng_is_setuid_setgid()) { |
255 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
256 | "-g, --group"); | |
257 | } else { | |
258 | tracing_group_name = strdup(arg); | |
259 | if (tracing_group_name == NULL) { | |
260 | ret = -errno; | |
261 | PERROR("strdup"); | |
262 | goto end; | |
263 | } | |
264 | tracing_group_name_override = 1; | |
330a40bb | 265 | } |
cd60b05a JG |
266 | break; |
267 | case 'h': | |
268 | usage(); | |
269 | exit(EXIT_FAILURE); | |
270 | case 'o': | |
e8fa9fb0 MD |
271 | if (lttng_is_setuid_setgid()) { |
272 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
273 | "-o, --output"); | |
274 | } else { | |
275 | ret = asprintf(&opt_output_path, "%s", arg); | |
276 | if (ret < 0) { | |
277 | ret = -errno; | |
278 | PERROR("asprintf opt_output_path"); | |
279 | goto end; | |
280 | } | |
cd60b05a JG |
281 | } |
282 | break; | |
283 | case 'v': | |
284 | /* Verbose level can increase using multiple -v */ | |
285 | if (arg) { | |
286 | lttng_opt_verbose = config_parse_value(arg); | |
287 | } else { | |
849e5b7b DG |
288 | /* Only 3 level of verbosity (-vvv). */ |
289 | if (lttng_opt_verbose < 3) { | |
290 | lttng_opt_verbose += 1; | |
291 | } | |
cd60b05a JG |
292 | } |
293 | break; | |
294 | default: | |
295 | /* Unknown option or other error. | |
296 | * Error is printed by getopt, just return */ | |
297 | ret = -1; | |
298 | goto end; | |
299 | } | |
300 | ||
301 | /* All good. */ | |
302 | ret = 0; | |
303 | ||
304 | end: | |
305 | return ret; | |
306 | } | |
307 | ||
308 | /* | |
309 | * config_entry_handler_cb used to handle options read from a config file. | |
310 | * See config_entry_handler_cb comment in common/config/config.h for the | |
311 | * return value conventions. | |
312 | */ | |
5a693d30 | 313 | static int config_entry_handler(const struct config_entry *entry, void *unused) |
cd60b05a JG |
314 | { |
315 | int ret = 0, i; | |
316 | ||
317 | if (!entry || !entry->name || !entry->value) { | |
318 | ret = -EINVAL; | |
319 | goto end; | |
320 | } | |
321 | ||
322 | /* Check if the option is to be ignored */ | |
323 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
324 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
325 | goto end; | |
326 | } | |
327 | } | |
328 | ||
329 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
330 | /* Ignore if entry name is not fully matched. */ | |
331 | if (strcmp(entry->name, long_options[i].name)) { | |
332 | continue; | |
333 | } | |
334 | ||
335 | /* | |
5a693d30 MD |
336 | * If the option takes no argument on the command line, |
337 | * we have to check if the value is "true". We support | |
338 | * non-zero numeric values, true, on and yes. | |
cd60b05a JG |
339 | */ |
340 | if (!long_options[i].has_arg) { | |
341 | ret = config_parse_value(entry->value); | |
342 | if (ret <= 0) { | |
343 | if (ret) { | |
344 | WARN("Invalid configuration value \"%s\" for option %s", | |
345 | entry->value, entry->name); | |
346 | } | |
347 | /* False, skip boolean config option. */ | |
348 | goto end; | |
349 | } | |
350 | } | |
351 | ||
352 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
353 | goto end; | |
354 | } | |
355 | ||
356 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
357 | entry->name); | |
358 | ||
359 | end: | |
360 | return ret; | |
361 | } | |
362 | ||
5a693d30 | 363 | static int set_options(int argc, char **argv) |
cd60b05a | 364 | { |
178a0557 | 365 | int c, ret = 0, option_index = 0, retval = 0; |
cd60b05a JG |
366 | int orig_optopt = optopt, orig_optind = optind; |
367 | char *default_address, *optstring; | |
368 | const char *config_path = NULL; | |
369 | ||
370 | optstring = utils_generate_optstring(long_options, | |
371 | sizeof(long_options) / sizeof(struct option)); | |
372 | if (!optstring) { | |
178a0557 | 373 | retval = -ENOMEM; |
cd60b05a JG |
374 | goto exit; |
375 | } | |
376 | ||
377 | /* Check for the --config option */ | |
378 | ||
379 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
380 | &option_index)) != -1) { | |
381 | if (c == '?') { | |
178a0557 | 382 | retval = -EINVAL; |
cd60b05a JG |
383 | goto exit; |
384 | } else if (c != 'f') { | |
385 | continue; | |
386 | } | |
387 | ||
e8fa9fb0 MD |
388 | if (lttng_is_setuid_setgid()) { |
389 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
390 | "-f, --config"); | |
391 | } else { | |
392 | config_path = utils_expand_path(optarg); | |
393 | if (!config_path) { | |
394 | ERR("Failed to resolve path: %s", optarg); | |
395 | } | |
cd60b05a JG |
396 | } |
397 | } | |
398 | ||
399 | ret = config_get_section_entries(config_path, config_section_name, | |
400 | config_entry_handler, NULL); | |
401 | if (ret) { | |
402 | if (ret > 0) { | |
403 | ERR("Invalid configuration option at line %i", ret); | |
cd60b05a | 404 | } |
178a0557 | 405 | retval = -1; |
cd60b05a JG |
406 | goto exit; |
407 | } | |
b8aa1682 | 408 | |
cd60b05a JG |
409 | /* Reset getopt's global state */ |
410 | optopt = orig_optopt; | |
411 | optind = orig_optind; | |
b8aa1682 | 412 | while (1) { |
cd60b05a | 413 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
414 | if (c == -1) { |
415 | break; | |
416 | } | |
417 | ||
cd60b05a JG |
418 | ret = set_option(c, optarg, long_options[option_index].name); |
419 | if (ret < 0) { | |
178a0557 | 420 | retval = -1; |
b8aa1682 JD |
421 | goto exit; |
422 | } | |
423 | } | |
424 | ||
425 | /* assign default values */ | |
426 | if (control_uri == NULL) { | |
fa91dc52 MD |
427 | ret = asprintf(&default_address, |
428 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", | |
429 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
430 | if (ret < 0) { |
431 | PERROR("asprintf default data address"); | |
178a0557 | 432 | retval = -1; |
b8aa1682 JD |
433 | goto exit; |
434 | } | |
435 | ||
436 | ret = uri_parse(default_address, &control_uri); | |
437 | free(default_address); | |
438 | if (ret < 0) { | |
439 | ERR("Invalid control URI specified"); | |
178a0557 | 440 | retval = -1; |
b8aa1682 JD |
441 | goto exit; |
442 | } | |
443 | } | |
444 | if (data_uri == NULL) { | |
fa91dc52 MD |
445 | ret = asprintf(&default_address, |
446 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", | |
447 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
448 | if (ret < 0) { |
449 | PERROR("asprintf default data address"); | |
178a0557 | 450 | retval = -1; |
b8aa1682 JD |
451 | goto exit; |
452 | } | |
453 | ||
454 | ret = uri_parse(default_address, &data_uri); | |
455 | free(default_address); | |
456 | if (ret < 0) { | |
457 | ERR("Invalid data URI specified"); | |
178a0557 | 458 | retval = -1; |
b8aa1682 JD |
459 | goto exit; |
460 | } | |
461 | } | |
d3e2ba59 | 462 | if (live_uri == NULL) { |
fa91dc52 MD |
463 | ret = asprintf(&default_address, |
464 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", | |
465 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
466 | if (ret < 0) { |
467 | PERROR("asprintf default viewer control address"); | |
178a0557 | 468 | retval = -1; |
d3e2ba59 JD |
469 | goto exit; |
470 | } | |
471 | ||
472 | ret = uri_parse(default_address, &live_uri); | |
473 | free(default_address); | |
474 | if (ret < 0) { | |
475 | ERR("Invalid viewer control URI specified"); | |
178a0557 | 476 | retval = -1; |
d3e2ba59 JD |
477 | goto exit; |
478 | } | |
479 | } | |
b8aa1682 JD |
480 | |
481 | exit: | |
cd60b05a | 482 | free(optstring); |
178a0557 | 483 | return retval; |
b8aa1682 JD |
484 | } |
485 | ||
5a693d30 MD |
486 | static void print_global_objects(void) |
487 | { | |
488 | rcu_register_thread(); | |
489 | ||
490 | print_viewer_streams(); | |
491 | print_relay_streams(); | |
492 | print_sessions(); | |
493 | ||
494 | rcu_unregister_thread(); | |
495 | } | |
496 | ||
b8aa1682 JD |
497 | /* |
498 | * Cleanup the daemon | |
499 | */ | |
5a693d30 | 500 | static void relayd_cleanup(void) |
b8aa1682 | 501 | { |
5a693d30 MD |
502 | print_global_objects(); |
503 | ||
b8aa1682 JD |
504 | DBG("Cleaning up"); |
505 | ||
178a0557 MD |
506 | if (viewer_streams_ht) |
507 | lttng_ht_destroy(viewer_streams_ht); | |
508 | if (relay_streams_ht) | |
509 | lttng_ht_destroy(relay_streams_ht); | |
5a693d30 MD |
510 | if (sessions_ht) |
511 | lttng_ht_destroy(sessions_ht); | |
178a0557 | 512 | |
095a4ae5 MD |
513 | /* free the dynamically allocated opt_output_path */ |
514 | free(opt_output_path); | |
515 | ||
a02de639 CB |
516 | /* Close thread quit pipes */ |
517 | utils_close_pipe(thread_quit_pipe); | |
518 | ||
710c1f73 DG |
519 | uri_free(control_uri); |
520 | uri_free(data_uri); | |
8d5c808e | 521 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
522 | |
523 | if (tracing_group_name_override) { | |
524 | free((void *) tracing_group_name); | |
525 | } | |
b8aa1682 JD |
526 | } |
527 | ||
528 | /* | |
529 | * Write to writable pipe used to notify a thread. | |
530 | */ | |
5a693d30 | 531 | static int notify_thread_pipe(int wpipe) |
b8aa1682 | 532 | { |
6cd525e8 | 533 | ssize_t ret; |
b8aa1682 | 534 | |
6cd525e8 MD |
535 | ret = lttng_write(wpipe, "!", 1); |
536 | if (ret < 1) { | |
b8aa1682 | 537 | PERROR("write poll pipe"); |
b4aacfdc | 538 | goto end; |
b8aa1682 | 539 | } |
b4aacfdc MD |
540 | ret = 0; |
541 | end: | |
b8aa1682 JD |
542 | return ret; |
543 | } | |
544 | ||
5a693d30 | 545 | static int notify_health_quit_pipe(int *pipe) |
65931c8b | 546 | { |
6cd525e8 | 547 | ssize_t ret; |
65931c8b | 548 | |
6cd525e8 MD |
549 | ret = lttng_write(pipe[1], "4", 1); |
550 | if (ret < 1) { | |
65931c8b | 551 | PERROR("write relay health quit"); |
b4aacfdc | 552 | goto end; |
65931c8b | 553 | } |
b4aacfdc MD |
554 | ret = 0; |
555 | end: | |
556 | return ret; | |
65931c8b MD |
557 | } |
558 | ||
b8aa1682 | 559 | /* |
b4aacfdc | 560 | * Stop all relayd and relayd-live threads. |
b8aa1682 | 561 | */ |
b4aacfdc | 562 | int lttng_relay_stop_threads(void) |
b8aa1682 | 563 | { |
b4aacfdc | 564 | int retval = 0; |
b8aa1682 JD |
565 | |
566 | /* Stopping all threads */ | |
567 | DBG("Terminating all threads"); | |
b4aacfdc | 568 | if (notify_thread_pipe(thread_quit_pipe[1])) { |
b8aa1682 | 569 | ERR("write error on thread quit pipe"); |
b4aacfdc | 570 | retval = -1; |
b8aa1682 JD |
571 | } |
572 | ||
b4aacfdc MD |
573 | if (notify_health_quit_pipe(health_quit_pipe)) { |
574 | ERR("write error on health quit pipe"); | |
575 | } | |
65931c8b | 576 | |
b8aa1682 | 577 | /* Dispatch thread */ |
26c9d55e | 578 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 579 | futex_nto1_wake(&relay_conn_queue.futex); |
178a0557 | 580 | |
b4aacfdc | 581 | if (relayd_live_stop()) { |
178a0557 | 582 | ERR("Error stopping live threads"); |
b4aacfdc | 583 | retval = -1; |
178a0557 | 584 | } |
b4aacfdc | 585 | return retval; |
b8aa1682 JD |
586 | } |
587 | ||
588 | /* | |
589 | * Signal handler for the daemon | |
590 | * | |
591 | * Simply stop all worker threads, leaving main() return gracefully after | |
592 | * joining all threads and calling cleanup(). | |
593 | */ | |
5a693d30 | 594 | static void sighandler(int sig) |
b8aa1682 JD |
595 | { |
596 | switch (sig) { | |
597 | case SIGPIPE: | |
598 | DBG("SIGPIPE caught"); | |
599 | return; | |
600 | case SIGINT: | |
601 | DBG("SIGINT caught"); | |
b4aacfdc MD |
602 | if (lttng_relay_stop_threads()) { |
603 | ERR("Error stopping threads"); | |
604 | } | |
b8aa1682 JD |
605 | break; |
606 | case SIGTERM: | |
607 | DBG("SIGTERM caught"); | |
b4aacfdc MD |
608 | if (lttng_relay_stop_threads()) { |
609 | ERR("Error stopping threads"); | |
610 | } | |
b8aa1682 | 611 | break; |
3fd27398 MD |
612 | case SIGUSR1: |
613 | CMM_STORE_SHARED(recv_child_signal, 1); | |
614 | break; | |
b8aa1682 JD |
615 | default: |
616 | break; | |
617 | } | |
618 | } | |
619 | ||
620 | /* | |
621 | * Setup signal handler for : | |
622 | * SIGINT, SIGTERM, SIGPIPE | |
623 | */ | |
5a693d30 | 624 | static int set_signal_handler(void) |
b8aa1682 JD |
625 | { |
626 | int ret = 0; | |
627 | struct sigaction sa; | |
628 | sigset_t sigset; | |
629 | ||
630 | if ((ret = sigemptyset(&sigset)) < 0) { | |
631 | PERROR("sigemptyset"); | |
632 | return ret; | |
633 | } | |
634 | ||
635 | sa.sa_handler = sighandler; | |
636 | sa.sa_mask = sigset; | |
637 | sa.sa_flags = 0; | |
638 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { | |
639 | PERROR("sigaction"); | |
640 | return ret; | |
641 | } | |
642 | ||
643 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
644 | PERROR("sigaction"); | |
645 | return ret; | |
646 | } | |
647 | ||
648 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
649 | PERROR("sigaction"); | |
650 | return ret; | |
651 | } | |
652 | ||
3fd27398 MD |
653 | if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { |
654 | PERROR("sigaction"); | |
655 | return ret; | |
656 | } | |
657 | ||
658 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
659 | |
660 | return ret; | |
661 | } | |
662 | ||
3fd27398 MD |
663 | void lttng_relay_notify_ready(void) |
664 | { | |
665 | /* Notify the parent of the fork() process that we are ready. */ | |
666 | if (opt_daemon || opt_background) { | |
667 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
668 | kill(child_ppid, SIGUSR1); | |
669 | } | |
670 | } | |
671 | } | |
672 | ||
b8aa1682 JD |
673 | /* |
674 | * Init thread quit pipe. | |
675 | * | |
676 | * Return -1 on error or 0 if all pipes are created. | |
677 | */ | |
5a693d30 | 678 | static int init_thread_quit_pipe(void) |
b8aa1682 | 679 | { |
a02de639 | 680 | int ret; |
b8aa1682 | 681 | |
a02de639 | 682 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 683 | |
b8aa1682 JD |
684 | return ret; |
685 | } | |
686 | ||
687 | /* | |
688 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
689 | */ | |
5a693d30 | 690 | static int create_thread_poll_set(struct lttng_poll_event *events, int size) |
b8aa1682 JD |
691 | { |
692 | int ret; | |
693 | ||
694 | if (events == NULL || size == 0) { | |
695 | ret = -1; | |
696 | goto error; | |
697 | } | |
698 | ||
699 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
700 | if (ret < 0) { | |
701 | goto error; | |
702 | } | |
703 | ||
704 | /* Add quit pipe */ | |
c7759e6a | 705 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR); |
b8aa1682 JD |
706 | if (ret < 0) { |
707 | goto error; | |
708 | } | |
709 | ||
710 | return 0; | |
711 | ||
712 | error: | |
713 | return ret; | |
714 | } | |
715 | ||
716 | /* | |
717 | * Check if the thread quit pipe was triggered. | |
718 | * | |
719 | * Return 1 if it was triggered else 0; | |
720 | */ | |
5a693d30 | 721 | static int check_thread_quit_pipe(int fd, uint32_t events) |
b8aa1682 JD |
722 | { |
723 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
724 | return 1; | |
725 | } | |
726 | ||
727 | return 0; | |
728 | } | |
729 | ||
730 | /* | |
731 | * Create and init socket from uri. | |
732 | */ | |
5a693d30 | 733 | static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri) |
b8aa1682 JD |
734 | { |
735 | int ret; | |
736 | struct lttcomm_sock *sock = NULL; | |
737 | ||
738 | sock = lttcomm_alloc_sock_from_uri(uri); | |
739 | if (sock == NULL) { | |
740 | ERR("Allocating socket"); | |
741 | goto error; | |
742 | } | |
743 | ||
744 | ret = lttcomm_create_sock(sock); | |
745 | if (ret < 0) { | |
746 | goto error; | |
747 | } | |
748 | DBG("Listening on sock %d", sock->fd); | |
749 | ||
750 | ret = sock->ops->bind(sock); | |
751 | if (ret < 0) { | |
752 | goto error; | |
753 | } | |
754 | ||
755 | ret = sock->ops->listen(sock, -1); | |
756 | if (ret < 0) { | |
757 | goto error; | |
758 | ||
759 | } | |
760 | ||
761 | return sock; | |
762 | ||
763 | error: | |
764 | if (sock) { | |
765 | lttcomm_destroy_sock(sock); | |
766 | } | |
767 | return NULL; | |
768 | } | |
769 | ||
770 | /* | |
771 | * This thread manages the listening for new connections on the network | |
772 | */ | |
5a693d30 | 773 | static void *relay_thread_listener(void *data) |
b8aa1682 | 774 | { |
095a4ae5 | 775 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
776 | uint32_t revents, nb_fd; |
777 | struct lttng_poll_event events; | |
778 | struct lttcomm_sock *control_sock, *data_sock; | |
779 | ||
b8aa1682 JD |
780 | DBG("[thread] Relay listener started"); |
781 | ||
55706a7d MD |
782 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
783 | ||
f385ae0a MD |
784 | health_code_update(); |
785 | ||
5a693d30 | 786 | control_sock = relay_socket_create(control_uri); |
b8aa1682 | 787 | if (!control_sock) { |
095a4ae5 | 788 | goto error_sock_control; |
b8aa1682 JD |
789 | } |
790 | ||
5a693d30 | 791 | data_sock = relay_socket_create(data_uri); |
b8aa1682 | 792 | if (!data_sock) { |
095a4ae5 | 793 | goto error_sock_relay; |
b8aa1682 JD |
794 | } |
795 | ||
796 | /* | |
5a693d30 MD |
797 | * Pass 3 as size here for the thread quit pipe, control and |
798 | * data socket. | |
b8aa1682 JD |
799 | */ |
800 | ret = create_thread_poll_set(&events, 3); | |
801 | if (ret < 0) { | |
802 | goto error_create_poll; | |
803 | } | |
804 | ||
805 | /* Add the control socket */ | |
806 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
807 | if (ret < 0) { | |
808 | goto error_poll_add; | |
809 | } | |
810 | ||
811 | /* Add the data socket */ | |
812 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
813 | if (ret < 0) { | |
814 | goto error_poll_add; | |
815 | } | |
816 | ||
3fd27398 MD |
817 | lttng_relay_notify_ready(); |
818 | ||
9b5e0863 MD |
819 | if (testpoint(relayd_thread_listener)) { |
820 | goto error_testpoint; | |
821 | } | |
822 | ||
b8aa1682 | 823 | while (1) { |
f385ae0a MD |
824 | health_code_update(); |
825 | ||
b8aa1682 JD |
826 | DBG("Listener accepting connections"); |
827 | ||
b8aa1682 | 828 | restart: |
f385ae0a | 829 | health_poll_entry(); |
b8aa1682 | 830 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 831 | health_poll_exit(); |
b8aa1682 JD |
832 | if (ret < 0) { |
833 | /* | |
834 | * Restart interrupted system call. | |
835 | */ | |
836 | if (errno == EINTR) { | |
837 | goto restart; | |
838 | } | |
839 | goto error; | |
840 | } | |
841 | ||
0d9c5d77 DG |
842 | nb_fd = ret; |
843 | ||
b8aa1682 JD |
844 | DBG("Relay new connection received"); |
845 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
846 | health_code_update(); |
847 | ||
b8aa1682 JD |
848 | /* Fetch once the poll data */ |
849 | revents = LTTNG_POLL_GETEV(&events, i); | |
850 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
851 | ||
fd20dac9 | 852 | if (!revents) { |
5a693d30 MD |
853 | /* |
854 | * No activity for this FD (poll | |
855 | * implementation). | |
856 | */ | |
fd20dac9 MD |
857 | continue; |
858 | } | |
859 | ||
b8aa1682 JD |
860 | /* Thread quit pipe has been closed. Killing thread. */ |
861 | ret = check_thread_quit_pipe(pollfd, revents); | |
862 | if (ret) { | |
095a4ae5 MD |
863 | err = 0; |
864 | goto exit; | |
b8aa1682 JD |
865 | } |
866 | ||
92d6debb | 867 | if (revents & LPOLLIN) { |
4b7f17b2 | 868 | /* |
5a693d30 MD |
869 | * A new connection is requested, therefore a |
870 | * sessiond/consumerd connection is allocated in | |
871 | * this thread, enqueued to a global queue and | |
872 | * dequeued (and freed) in the worker thread. | |
4b7f17b2 | 873 | */ |
58eb9381 DG |
874 | int val = 1; |
875 | struct relay_connection *new_conn; | |
4b7f17b2 | 876 | struct lttcomm_sock *newsock; |
5a693d30 | 877 | enum connection_type type; |
b8aa1682 JD |
878 | |
879 | if (pollfd == data_sock->fd) { | |
5a693d30 | 880 | type = RELAY_DATA; |
b8aa1682 | 881 | newsock = data_sock->ops->accept(data_sock); |
58eb9381 DG |
882 | DBG("Relay data connection accepted, socket %d", |
883 | newsock->fd); | |
4b7f17b2 MD |
884 | } else { |
885 | assert(pollfd == control_sock->fd); | |
5a693d30 | 886 | type = RELAY_CONTROL; |
b8aa1682 | 887 | newsock = control_sock->ops->accept(control_sock); |
58eb9381 DG |
888 | DBG("Relay control connection accepted, socket %d", |
889 | newsock->fd); | |
b8aa1682 | 890 | } |
58eb9381 DG |
891 | if (!newsock) { |
892 | PERROR("accepting sock"); | |
58eb9381 DG |
893 | goto error; |
894 | } | |
895 | ||
896 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, | |
897 | sizeof(val)); | |
b8aa1682 JD |
898 | if (ret < 0) { |
899 | PERROR("setsockopt inet"); | |
4b7f17b2 | 900 | lttcomm_destroy_sock(newsock); |
b8aa1682 JD |
901 | goto error; |
902 | } | |
5a693d30 MD |
903 | new_conn = connection_create(newsock, type); |
904 | if (!new_conn) { | |
905 | lttcomm_destroy_sock(newsock); | |
906 | goto error; | |
907 | } | |
58eb9381 DG |
908 | |
909 | /* Enqueue request for the dispatcher thread. */ | |
8bdee6e2 SM |
910 | cds_wfcq_enqueue(&relay_conn_queue.head, &relay_conn_queue.tail, |
911 | &new_conn->qnode); | |
b8aa1682 JD |
912 | |
913 | /* | |
5a693d30 MD |
914 | * Wake the dispatch queue futex. |
915 | * Implicit memory barrier with the | |
916 | * exchange in cds_wfcq_enqueue. | |
b8aa1682 | 917 | */ |
58eb9381 | 918 | futex_nto1_wake(&relay_conn_queue.futex); |
92d6debb MD |
919 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
920 | ERR("socket poll error"); | |
921 | goto error; | |
922 | } else { | |
923 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
924 | goto error; | |
b8aa1682 JD |
925 | } |
926 | } | |
927 | } | |
928 | ||
095a4ae5 | 929 | exit: |
b8aa1682 JD |
930 | error: |
931 | error_poll_add: | |
9b5e0863 | 932 | error_testpoint: |
b8aa1682 JD |
933 | lttng_poll_clean(&events); |
934 | error_create_poll: | |
095a4ae5 MD |
935 | if (data_sock->fd >= 0) { |
936 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
937 | if (ret) { |
938 | PERROR("close"); | |
939 | } | |
b8aa1682 | 940 | } |
095a4ae5 MD |
941 | lttcomm_destroy_sock(data_sock); |
942 | error_sock_relay: | |
943 | if (control_sock->fd >= 0) { | |
944 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
945 | if (ret) { |
946 | PERROR("close"); | |
947 | } | |
b8aa1682 | 948 | } |
095a4ae5 MD |
949 | lttcomm_destroy_sock(control_sock); |
950 | error_sock_control: | |
951 | if (err) { | |
f385ae0a MD |
952 | health_error(); |
953 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 954 | } |
55706a7d | 955 | health_unregister(health_relayd); |
b8aa1682 | 956 | DBG("Relay listener thread cleanup complete"); |
b4aacfdc | 957 | lttng_relay_stop_threads(); |
b8aa1682 JD |
958 | return NULL; |
959 | } | |
960 | ||
961 | /* | |
962 | * This thread manages the dispatching of the requests to worker threads | |
963 | */ | |
5a693d30 | 964 | static void *relay_thread_dispatcher(void *data) |
b8aa1682 | 965 | { |
6cd525e8 MD |
966 | int err = -1; |
967 | ssize_t ret; | |
8bdee6e2 | 968 | struct cds_wfcq_node *node; |
58eb9381 | 969 | struct relay_connection *new_conn = NULL; |
b8aa1682 JD |
970 | |
971 | DBG("[thread] Relay dispatcher started"); | |
972 | ||
55706a7d MD |
973 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
974 | ||
9b5e0863 MD |
975 | if (testpoint(relayd_thread_dispatcher)) { |
976 | goto error_testpoint; | |
977 | } | |
978 | ||
f385ae0a MD |
979 | health_code_update(); |
980 | ||
26c9d55e | 981 | while (!CMM_LOAD_SHARED(dispatch_thread_exit)) { |
f385ae0a MD |
982 | health_code_update(); |
983 | ||
b8aa1682 | 984 | /* Atomically prepare the queue futex */ |
58eb9381 | 985 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 JD |
986 | |
987 | do { | |
f385ae0a MD |
988 | health_code_update(); |
989 | ||
b8aa1682 | 990 | /* Dequeue commands */ |
8bdee6e2 SM |
991 | node = cds_wfcq_dequeue_blocking(&relay_conn_queue.head, |
992 | &relay_conn_queue.tail); | |
b8aa1682 JD |
993 | if (node == NULL) { |
994 | DBG("Woken up but nothing in the relay command queue"); | |
995 | /* Continue thread execution */ | |
996 | break; | |
997 | } | |
58eb9381 | 998 | new_conn = caa_container_of(node, struct relay_connection, qnode); |
b8aa1682 | 999 | |
58eb9381 | 1000 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
1001 | |
1002 | /* | |
5a693d30 MD |
1003 | * Inform worker thread of the new request. This |
1004 | * call is blocking so we can be assured that | |
1005 | * the data will be read at some point in time | |
1006 | * or wait to the end of the world :) | |
b8aa1682 | 1007 | */ |
58eb9381 DG |
1008 | ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn)); |
1009 | if (ret < 0) { | |
1010 | PERROR("write connection pipe"); | |
5a693d30 | 1011 | connection_put(new_conn); |
b8aa1682 JD |
1012 | goto error; |
1013 | } | |
1014 | } while (node != NULL); | |
1015 | ||
1016 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 1017 | health_poll_entry(); |
58eb9381 | 1018 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 1019 | health_poll_exit(); |
b8aa1682 JD |
1020 | } |
1021 | ||
f385ae0a MD |
1022 | /* Normal exit, no error */ |
1023 | err = 0; | |
1024 | ||
b8aa1682 | 1025 | error: |
9b5e0863 | 1026 | error_testpoint: |
f385ae0a MD |
1027 | if (err) { |
1028 | health_error(); | |
1029 | ERR("Health error occurred in %s", __func__); | |
1030 | } | |
55706a7d | 1031 | health_unregister(health_relayd); |
b8aa1682 | 1032 | DBG("Dispatch thread dying"); |
b4aacfdc | 1033 | lttng_relay_stop_threads(); |
b8aa1682 JD |
1034 | return NULL; |
1035 | } | |
1036 | ||
b8aa1682 | 1037 | /* |
5a693d30 | 1038 | * Set index data from the control port to a given index object. |
b8aa1682 | 1039 | */ |
5a693d30 | 1040 | static int set_index_control_data(struct relay_index *index, |
1c20f0e2 JD |
1041 | struct lttcomm_relayd_index *data) |
1042 | { | |
5a693d30 | 1043 | struct ctf_packet_index index_data; |
1c20f0e2 JD |
1044 | |
1045 | /* | |
5a693d30 MD |
1046 | * The index on disk is encoded in big endian, so we don't need |
1047 | * to convert the data received on the network. The data_offset | |
1048 | * value is NEVER modified here and is updated by the data | |
1049 | * thread. | |
1c20f0e2 | 1050 | */ |
5a693d30 MD |
1051 | index_data.packet_size = data->packet_size; |
1052 | index_data.content_size = data->content_size; | |
1053 | index_data.timestamp_begin = data->timestamp_begin; | |
1054 | index_data.timestamp_end = data->timestamp_end; | |
1055 | index_data.events_discarded = data->events_discarded; | |
1056 | index_data.stream_id = data->stream_id; | |
1057 | return relay_index_set_data(index, &index_data); | |
1c20f0e2 JD |
1058 | } |
1059 | ||
c5b6f4f0 DG |
1060 | /* |
1061 | * Handle the RELAYD_CREATE_SESSION command. | |
1062 | * | |
1063 | * On success, send back the session id or else return a negative value. | |
1064 | */ | |
5a693d30 | 1065 | static int relay_create_session(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1066 | struct relay_connection *conn) |
c5b6f4f0 DG |
1067 | { |
1068 | int ret = 0, send_ret; | |
1069 | struct relay_session *session; | |
1070 | struct lttcomm_relayd_status_session reply; | |
5a693d30 MD |
1071 | char session_name[NAME_MAX]; |
1072 | char hostname[HOST_NAME_MAX]; | |
1073 | uint32_t live_timer = 0; | |
1074 | bool snapshot = false; | |
c5b6f4f0 | 1075 | |
5a693d30 MD |
1076 | memset(session_name, 0, NAME_MAX); |
1077 | memset(hostname, 0, HOST_NAME_MAX); | |
c5b6f4f0 DG |
1078 | |
1079 | memset(&reply, 0, sizeof(reply)); | |
1080 | ||
58eb9381 | 1081 | switch (conn->minor) { |
2a174661 DG |
1082 | case 1: |
1083 | case 2: | |
1084 | case 3: | |
1085 | break; | |
1086 | case 4: /* LTTng sessiond 2.4 */ | |
1087 | default: | |
5a693d30 MD |
1088 | ret = cmd_create_session_2_4(conn, session_name, |
1089 | hostname, &live_timer, &snapshot); | |
1090 | } | |
1091 | if (ret < 0) { | |
1092 | goto send_reply; | |
d3e2ba59 JD |
1093 | } |
1094 | ||
5a693d30 MD |
1095 | session = session_create(session_name, hostname, live_timer, |
1096 | snapshot, conn->major, conn->minor); | |
1097 | if (!session) { | |
1098 | ret = -1; | |
1099 | goto send_reply; | |
1100 | } | |
1101 | assert(!conn->session); | |
1102 | conn->session = session; | |
c5b6f4f0 DG |
1103 | DBG("Created session %" PRIu64, session->id); |
1104 | ||
5a693d30 MD |
1105 | reply.session_id = htobe64(session->id); |
1106 | ||
1107 | send_reply: | |
c5b6f4f0 DG |
1108 | if (ret < 0) { |
1109 | reply.ret_code = htobe32(LTTNG_ERR_FATAL); | |
1110 | } else { | |
1111 | reply.ret_code = htobe32(LTTNG_OK); | |
1112 | } | |
1113 | ||
58eb9381 | 1114 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c5b6f4f0 DG |
1115 | if (send_ret < 0) { |
1116 | ERR("Relayd sending session id"); | |
4169f5ad | 1117 | ret = send_ret; |
c5b6f4f0 DG |
1118 | } |
1119 | ||
1120 | return ret; | |
1121 | } | |
1122 | ||
a4baae1b JD |
1123 | /* |
1124 | * When we have received all the streams and the metadata for a channel, | |
1125 | * we make them visible to the viewer threads. | |
1126 | */ | |
5a693d30 | 1127 | static void publish_connection_local_streams(struct relay_connection *conn) |
a4baae1b | 1128 | { |
5a693d30 MD |
1129 | struct relay_stream *stream; |
1130 | struct relay_session *session = conn->session; | |
a4baae1b | 1131 | |
5a693d30 MD |
1132 | /* |
1133 | * We publish all streams belonging to a session atomically wrt | |
1134 | * session lock. | |
1135 | */ | |
1136 | pthread_mutex_lock(&session->lock); | |
1137 | rcu_read_lock(); | |
1138 | cds_list_for_each_entry_rcu(stream, &session->recv_list, | |
1139 | recv_node) { | |
1140 | stream_publish(stream); | |
a4baae1b | 1141 | } |
5a693d30 | 1142 | rcu_read_unlock(); |
a4baae1b | 1143 | |
5a693d30 MD |
1144 | /* |
1145 | * Inform the viewer that there are new streams in the session. | |
1146 | */ | |
1147 | if (session->viewer_attached) { | |
1148 | uatomic_set(&session->new_streams, 1); | |
1149 | } | |
1150 | pthread_mutex_unlock(&session->lock); | |
a4baae1b JD |
1151 | } |
1152 | ||
b8aa1682 JD |
1153 | /* |
1154 | * relay_add_stream: allocate a new stream for a session | |
1155 | */ | |
5a693d30 | 1156 | static int relay_add_stream(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1157 | struct relay_connection *conn) |
b8aa1682 | 1158 | { |
5a693d30 MD |
1159 | int ret; |
1160 | ssize_t send_ret; | |
58eb9381 | 1161 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1162 | struct relay_stream *stream = NULL; |
1163 | struct lttcomm_relayd_status_stream reply; | |
80ec61a6 | 1164 | struct ctf_trace *trace = NULL; |
5a693d30 MD |
1165 | uint64_t stream_handle = -1ULL; |
1166 | char *path_name = NULL, *channel_name = NULL; | |
1167 | uint64_t tracefile_size = 0, tracefile_count = 0; | |
b8aa1682 | 1168 | |
58eb9381 | 1169 | if (!session || conn->version_check_done == 0) { |
b8aa1682 JD |
1170 | ERR("Trying to add a stream before version check"); |
1171 | ret = -1; | |
1172 | goto end_no_session; | |
1173 | } | |
1174 | ||
5a693d30 MD |
1175 | switch (session->minor) { |
1176 | case 1: /* LTTng sessiond 2.1. Allocates path_name and channel_name. */ | |
1177 | ret = cmd_recv_stream_2_1(conn, &path_name, | |
1178 | &channel_name); | |
0f907de1 | 1179 | break; |
5a693d30 | 1180 | case 2: /* LTTng sessiond 2.2. Allocates path_name and channel_name. */ |
0f907de1 | 1181 | default: |
5a693d30 MD |
1182 | ret = cmd_recv_stream_2_2(conn, &path_name, |
1183 | &channel_name, &tracefile_size, &tracefile_count); | |
0f907de1 JD |
1184 | break; |
1185 | } | |
1186 | if (ret < 0) { | |
5a693d30 | 1187 | goto send_reply; |
b8aa1682 JD |
1188 | } |
1189 | ||
5a693d30 | 1190 | trace = ctf_trace_get_by_path_or_create(session, path_name); |
2a174661 | 1191 | if (!trace) { |
5a693d30 | 1192 | goto send_reply; |
2a174661 | 1193 | } |
5a693d30 | 1194 | /* This stream here has one reference on the trace. */ |
2a174661 | 1195 | |
5a693d30 MD |
1196 | pthread_mutex_lock(&last_relay_stream_id_lock); |
1197 | stream_handle = ++last_relay_stream_id; | |
1198 | pthread_mutex_unlock(&last_relay_stream_id_lock); | |
d3e2ba59 | 1199 | |
5a693d30 MD |
1200 | /* We pass ownership of path_name and channel_name. */ |
1201 | stream = stream_create(trace, stream_handle, path_name, | |
1202 | channel_name, tracefile_size, tracefile_count); | |
1203 | path_name = NULL; | |
1204 | channel_name = NULL; | |
a4baae1b | 1205 | |
2a174661 | 1206 | /* |
5a693d30 MD |
1207 | * Streams are the owners of their trace. Reference to trace is |
1208 | * kept within stream_create(). | |
2a174661 | 1209 | */ |
5a693d30 | 1210 | ctf_trace_put(trace); |
d3e2ba59 | 1211 | |
5a693d30 | 1212 | send_reply: |
53efb85a | 1213 | memset(&reply, 0, sizeof(reply)); |
5a693d30 MD |
1214 | reply.handle = htobe64(stream_handle); |
1215 | if (!stream) { | |
f73fabfd | 1216 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
b8aa1682 | 1217 | } else { |
f73fabfd | 1218 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1219 | } |
5af40280 | 1220 | |
58eb9381 | 1221 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1222 | sizeof(struct lttcomm_relayd_status_stream), 0); |
1223 | if (send_ret < 0) { | |
1224 | ERR("Relay sending stream id"); | |
5a693d30 | 1225 | ret = (int) send_ret; |
b8aa1682 JD |
1226 | } |
1227 | ||
1228 | end_no_session: | |
5a693d30 MD |
1229 | free(path_name); |
1230 | free(channel_name); | |
0f907de1 | 1231 | return ret; |
b8aa1682 JD |
1232 | } |
1233 | ||
173af62f DG |
1234 | /* |
1235 | * relay_close_stream: close a specific stream | |
1236 | */ | |
5a693d30 | 1237 | static int relay_close_stream(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1238 | struct relay_connection *conn) |
173af62f | 1239 | { |
94d49140 | 1240 | int ret, send_ret; |
58eb9381 | 1241 | struct relay_session *session = conn->session; |
173af62f DG |
1242 | struct lttcomm_relayd_close_stream stream_info; |
1243 | struct lttcomm_relayd_generic_reply reply; | |
1244 | struct relay_stream *stream; | |
173af62f DG |
1245 | |
1246 | DBG("Close stream received"); | |
1247 | ||
58eb9381 | 1248 | if (!session || conn->version_check_done == 0) { |
173af62f DG |
1249 | ERR("Trying to close a stream before version check"); |
1250 | ret = -1; | |
1251 | goto end_no_session; | |
1252 | } | |
1253 | ||
58eb9381 | 1254 | ret = conn->sock->ops->recvmsg(conn->sock, &stream_info, |
7c5aef62 | 1255 | sizeof(struct lttcomm_relayd_close_stream), 0); |
173af62f | 1256 | if (ret < sizeof(struct lttcomm_relayd_close_stream)) { |
a6cd2b97 DG |
1257 | if (ret == 0) { |
1258 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1259 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1260 | } else { |
1261 | ERR("Relay didn't receive valid add_stream struct size : %d", ret); | |
1262 | } | |
173af62f DG |
1263 | ret = -1; |
1264 | goto end_no_session; | |
1265 | } | |
1266 | ||
5a693d30 | 1267 | stream = stream_get_by_id(be64toh(stream_info.stream_id)); |
173af62f DG |
1268 | if (!stream) { |
1269 | ret = -1; | |
5a693d30 | 1270 | goto end; |
173af62f | 1271 | } |
a12886a5 MD |
1272 | |
1273 | /* | |
1274 | * Set last_net_seq_num before the close flag. Required by data | |
1275 | * pending check. | |
1276 | */ | |
5a693d30 | 1277 | pthread_mutex_lock(&stream->lock); |
8e2583a4 | 1278 | stream->last_net_seq_num = be64toh(stream_info.last_net_seq_num); |
a12886a5 MD |
1279 | pthread_mutex_unlock(&stream->lock); |
1280 | ||
a08107cc JG |
1281 | /* |
1282 | * This is one of the conditions which may trigger a stream close | |
1283 | * with the others being: | |
1284 | * 1) A close command is received for a stream | |
1285 | * 2) The control connection owning the stream is closed | |
1286 | * 3) We have received all of the stream's data _after_ a close | |
1287 | * request. | |
1288 | */ | |
1289 | try_stream_close(stream); | |
5a693d30 MD |
1290 | if (stream->is_metadata) { |
1291 | struct relay_viewer_stream *vstream; | |
173af62f | 1292 | |
5a693d30 MD |
1293 | vstream = viewer_stream_get_by_id(stream->stream_handle); |
1294 | if (vstream) { | |
1295 | if (vstream->metadata_sent == stream->metadata_received) { | |
1296 | /* | |
1297 | * Since all the metadata has been sent to the | |
1298 | * viewer and that we have a request to close | |
1299 | * its stream, we can safely teardown the | |
1300 | * corresponding metadata viewer stream. | |
1301 | */ | |
1302 | viewer_stream_put(vstream); | |
1303 | } | |
1304 | /* Put local reference. */ | |
1305 | viewer_stream_put(vstream); | |
1306 | } | |
1307 | } | |
5a693d30 | 1308 | stream_put(stream); |
173af62f | 1309 | |
5a693d30 | 1310 | end: |
53efb85a | 1311 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1312 | if (ret < 0) { |
f73fabfd | 1313 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1314 | } else { |
f73fabfd | 1315 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1316 | } |
58eb9381 | 1317 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
173af62f DG |
1318 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1319 | if (send_ret < 0) { | |
1320 | ERR("Relay sending stream id"); | |
4169f5ad | 1321 | ret = send_ret; |
173af62f DG |
1322 | } |
1323 | ||
1324 | end_no_session: | |
1325 | return ret; | |
1326 | } | |
1327 | ||
b8aa1682 JD |
1328 | /* |
1329 | * relay_unknown_command: send -1 if received unknown command | |
1330 | */ | |
5a693d30 | 1331 | static void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1332 | { |
1333 | struct lttcomm_relayd_generic_reply reply; | |
1334 | int ret; | |
1335 | ||
53efb85a | 1336 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1337 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
58eb9381 | 1338 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1339 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1340 | if (ret < 0) { | |
1341 | ERR("Relay sending unknown command"); | |
1342 | } | |
1343 | } | |
1344 | ||
1345 | /* | |
1346 | * relay_start: send an acknowledgment to the client to tell if we are | |
1347 | * ready to receive data. We are ready if a session is established. | |
1348 | */ | |
5a693d30 | 1349 | static int relay_start(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1350 | struct relay_connection *conn) |
b8aa1682 | 1351 | { |
f73fabfd | 1352 | int ret = htobe32(LTTNG_OK); |
b8aa1682 | 1353 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1354 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1355 | |
1356 | if (!session) { | |
1357 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1358 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1359 | } |
1360 | ||
53efb85a | 1361 | memset(&reply, 0, sizeof(reply)); |
b8aa1682 | 1362 | reply.ret_code = ret; |
58eb9381 | 1363 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1364 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1365 | if (ret < 0) { | |
1366 | ERR("Relay sending start ack"); | |
1367 | } | |
1368 | ||
1369 | return ret; | |
1370 | } | |
1371 | ||
1d4dfdef DG |
1372 | /* |
1373 | * Append padding to the file pointed by the file descriptor fd. | |
1374 | */ | |
1375 | static int write_padding_to_file(int fd, uint32_t size) | |
1376 | { | |
6cd525e8 | 1377 | ssize_t ret = 0; |
1d4dfdef DG |
1378 | char *zeros; |
1379 | ||
1380 | if (size == 0) { | |
1381 | goto end; | |
1382 | } | |
1383 | ||
1384 | zeros = zmalloc(size); | |
1385 | if (zeros == NULL) { | |
1386 | PERROR("zmalloc zeros for padding"); | |
1387 | ret = -1; | |
1388 | goto end; | |
1389 | } | |
1390 | ||
6cd525e8 MD |
1391 | ret = lttng_write(fd, zeros, size); |
1392 | if (ret < size) { | |
1d4dfdef DG |
1393 | PERROR("write padding to file"); |
1394 | } | |
1395 | ||
e986c7a1 DG |
1396 | free(zeros); |
1397 | ||
1d4dfdef DG |
1398 | end: |
1399 | return ret; | |
1400 | } | |
1401 | ||
b8aa1682 | 1402 | /* |
5a693d30 | 1403 | * relay_recv_metadata: receive the metadata for the session. |
b8aa1682 | 1404 | */ |
5a693d30 | 1405 | static int relay_recv_metadata(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1406 | struct relay_connection *conn) |
b8aa1682 | 1407 | { |
7fcdafc9 | 1408 | int ret = 0; |
6cd525e8 | 1409 | ssize_t size_ret; |
58eb9381 | 1410 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1411 | struct lttcomm_relayd_metadata_payload *metadata_struct; |
1412 | struct relay_stream *metadata_stream; | |
1413 | uint64_t data_size, payload_size; | |
1414 | ||
1415 | if (!session) { | |
1416 | ERR("Metadata sent before version check"); | |
1417 | ret = -1; | |
1418 | goto end; | |
1419 | } | |
1420 | ||
f6416125 MD |
1421 | data_size = payload_size = be64toh(recv_hdr->data_size); |
1422 | if (data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { | |
1423 | ERR("Incorrect data size"); | |
1424 | ret = -1; | |
1425 | goto end; | |
1426 | } | |
1427 | payload_size -= sizeof(struct lttcomm_relayd_metadata_payload); | |
1428 | ||
b8aa1682 | 1429 | if (data_buffer_size < data_size) { |
d7b3776f | 1430 | /* In case the realloc fails, we can free the memory */ |
c617c0c6 MD |
1431 | char *tmp_data_ptr; |
1432 | ||
1433 | tmp_data_ptr = realloc(data_buffer, data_size); | |
1434 | if (!tmp_data_ptr) { | |
b8aa1682 | 1435 | ERR("Allocating data buffer"); |
c617c0c6 | 1436 | free(data_buffer); |
b8aa1682 JD |
1437 | ret = -1; |
1438 | goto end; | |
1439 | } | |
c617c0c6 | 1440 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
1441 | data_buffer_size = data_size; |
1442 | } | |
1443 | memset(data_buffer, 0, data_size); | |
77c7c900 | 1444 | DBG2("Relay receiving metadata, waiting for %" PRIu64 " bytes", data_size); |
7fcdafc9 JG |
1445 | size_ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
1446 | if (size_ret < 0 || size_ret != data_size) { | |
1447 | if (size_ret == 0) { | |
a6cd2b97 | 1448 | /* Orderly shutdown. Not necessary to print an error. */ |
58eb9381 | 1449 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1450 | } else { |
1451 | ERR("Relay didn't receive the whole metadata"); | |
1452 | } | |
b8aa1682 | 1453 | ret = -1; |
b8aa1682 JD |
1454 | goto end; |
1455 | } | |
1456 | metadata_struct = (struct lttcomm_relayd_metadata_payload *) data_buffer; | |
9d1bbf21 | 1457 | |
5a693d30 | 1458 | metadata_stream = stream_get_by_id(be64toh(metadata_struct->stream_id)); |
b8aa1682 JD |
1459 | if (!metadata_stream) { |
1460 | ret = -1; | |
5a693d30 | 1461 | goto end; |
b8aa1682 JD |
1462 | } |
1463 | ||
5a693d30 MD |
1464 | pthread_mutex_lock(&metadata_stream->lock); |
1465 | ||
1466 | size_ret = lttng_write(metadata_stream->stream_fd->fd, metadata_struct->payload, | |
6cd525e8 MD |
1467 | payload_size); |
1468 | if (size_ret < payload_size) { | |
b8aa1682 JD |
1469 | ERR("Relay error writing metadata on file"); |
1470 | ret = -1; | |
5a693d30 | 1471 | goto end_put; |
b8aa1682 | 1472 | } |
1d4dfdef | 1473 | |
7fcdafc9 | 1474 | size_ret = write_padding_to_file(metadata_stream->stream_fd->fd, |
1d4dfdef | 1475 | be32toh(metadata_struct->padding_size)); |
7fcdafc9 | 1476 | if (size_ret < 0) { |
5a693d30 | 1477 | goto end_put; |
1d4dfdef | 1478 | } |
2a174661 | 1479 | |
5a693d30 | 1480 | metadata_stream->metadata_received += |
d3e2ba59 | 1481 | payload_size + be32toh(metadata_struct->padding_size); |
5a693d30 MD |
1482 | DBG2("Relay metadata written. Updated metadata_received %" PRIu64, |
1483 | metadata_stream->metadata_received); | |
1d4dfdef | 1484 | |
5a693d30 MD |
1485 | end_put: |
1486 | pthread_mutex_unlock(&metadata_stream->lock); | |
1487 | stream_put(metadata_stream); | |
b8aa1682 JD |
1488 | end: |
1489 | return ret; | |
1490 | } | |
1491 | ||
1492 | /* | |
1493 | * relay_send_version: send relayd version number | |
1494 | */ | |
5a693d30 | 1495 | static int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1496 | struct relay_connection *conn) |
b8aa1682 | 1497 | { |
7f51dcba | 1498 | int ret; |
092b6259 | 1499 | struct lttcomm_relayd_version reply, msg; |
b8aa1682 | 1500 | |
58eb9381 | 1501 | conn->version_check_done = 1; |
b8aa1682 | 1502 | |
092b6259 | 1503 | /* Get version from the other side. */ |
58eb9381 | 1504 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
092b6259 | 1505 | if (ret < 0 || ret != sizeof(msg)) { |
a6cd2b97 DG |
1506 | if (ret == 0) { |
1507 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1508 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1509 | } else { |
1510 | ERR("Relay failed to receive the version values."); | |
1511 | } | |
092b6259 | 1512 | ret = -1; |
092b6259 DG |
1513 | goto end; |
1514 | } | |
1515 | ||
53efb85a | 1516 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
1517 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1518 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1519 | |
1520 | /* Major versions must be the same */ | |
1521 | if (reply.major != be32toh(msg.major)) { | |
6151a90f JD |
1522 | DBG("Incompatible major versions (%u vs %u), deleting session", |
1523 | reply.major, be32toh(msg.major)); | |
5a693d30 | 1524 | connection_put(conn); |
d4519fa3 JD |
1525 | ret = 0; |
1526 | goto end; | |
1527 | } | |
1528 | ||
58eb9381 | 1529 | conn->major = reply.major; |
0f907de1 JD |
1530 | /* We adapt to the lowest compatible version */ |
1531 | if (reply.minor <= be32toh(msg.minor)) { | |
58eb9381 | 1532 | conn->minor = reply.minor; |
0f907de1 | 1533 | } else { |
58eb9381 | 1534 | conn->minor = be32toh(msg.minor); |
0f907de1 JD |
1535 | } |
1536 | ||
6151a90f JD |
1537 | reply.major = htobe32(reply.major); |
1538 | reply.minor = htobe32(reply.minor); | |
58eb9381 | 1539 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
6151a90f JD |
1540 | sizeof(struct lttcomm_relayd_version), 0); |
1541 | if (ret < 0) { | |
1542 | ERR("Relay sending version"); | |
1543 | } | |
1544 | ||
58eb9381 DG |
1545 | DBG("Version check done using protocol %u.%u", conn->major, |
1546 | conn->minor); | |
b8aa1682 JD |
1547 | |
1548 | end: | |
1549 | return ret; | |
1550 | } | |
1551 | ||
c8f59ee5 | 1552 | /* |
6d805429 | 1553 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 | 1554 | */ |
5a693d30 | 1555 | static int relay_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1556 | struct relay_connection *conn) |
c8f59ee5 | 1557 | { |
58eb9381 | 1558 | struct relay_session *session = conn->session; |
6d805429 | 1559 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1560 | struct lttcomm_relayd_generic_reply reply; |
1561 | struct relay_stream *stream; | |
1562 | int ret; | |
c8f59ee5 DG |
1563 | uint64_t last_net_seq_num, stream_id; |
1564 | ||
6d805429 | 1565 | DBG("Data pending command received"); |
c8f59ee5 | 1566 | |
58eb9381 | 1567 | if (!session || conn->version_check_done == 0) { |
c8f59ee5 DG |
1568 | ERR("Trying to check for data before version check"); |
1569 | ret = -1; | |
1570 | goto end_no_session; | |
1571 | } | |
1572 | ||
58eb9381 | 1573 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
c8f59ee5 | 1574 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1575 | if (ret == 0) { |
1576 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1577 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1578 | } else { |
1579 | ERR("Relay didn't receive valid data_pending struct size : %d", | |
1580 | ret); | |
1581 | } | |
c8f59ee5 DG |
1582 | ret = -1; |
1583 | goto end_no_session; | |
1584 | } | |
1585 | ||
1586 | stream_id = be64toh(msg.stream_id); | |
1587 | last_net_seq_num = be64toh(msg.last_net_seq_num); | |
1588 | ||
5a693d30 | 1589 | stream = stream_get_by_id(stream_id); |
de91f48a | 1590 | if (stream == NULL) { |
c8f59ee5 | 1591 | ret = -1; |
5a693d30 | 1592 | goto end; |
c8f59ee5 DG |
1593 | } |
1594 | ||
5a693d30 MD |
1595 | pthread_mutex_lock(&stream->lock); |
1596 | ||
6d805429 | 1597 | DBG("Data pending for stream id %" PRIu64 " prev_seq %" PRIu64 |
c8f59ee5 DG |
1598 | " and last_seq %" PRIu64, stream_id, stream->prev_seq, |
1599 | last_net_seq_num); | |
1600 | ||
33832e64 | 1601 | /* Avoid wrapping issue */ |
39df6d9f | 1602 | if (((int64_t) (stream->prev_seq - last_net_seq_num)) >= 0) { |
6d805429 | 1603 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1604 | ret = 0; |
6d805429 DG |
1605 | } else { |
1606 | /* Data still being streamed thus pending */ | |
1607 | ret = 1; | |
c8f59ee5 DG |
1608 | } |
1609 | ||
5a693d30 MD |
1610 | stream->data_pending_check_done = true; |
1611 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 | 1612 | |
5a693d30 MD |
1613 | stream_put(stream); |
1614 | end: | |
c8f59ee5 | 1615 | |
53efb85a | 1616 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1617 | reply.ret_code = htobe32(ret); |
58eb9381 | 1618 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1619 | if (ret < 0) { |
6d805429 | 1620 | ERR("Relay data pending ret code failed"); |
c8f59ee5 DG |
1621 | } |
1622 | ||
1623 | end_no_session: | |
1624 | return ret; | |
1625 | } | |
1626 | ||
1627 | /* | |
1628 | * Wait for the control socket to reach a quiescent state. | |
1629 | * | |
5a693d30 MD |
1630 | * Note that for now, when receiving this command from the session |
1631 | * daemon, this means that every subsequent commands or data received on | |
1632 | * the control socket has been handled. So, this is why we simply return | |
1633 | * OK here. | |
c8f59ee5 | 1634 | */ |
5a693d30 | 1635 | static int relay_quiescent_control(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1636 | struct relay_connection *conn) |
c8f59ee5 DG |
1637 | { |
1638 | int ret; | |
ad7051c0 DG |
1639 | uint64_t stream_id; |
1640 | struct relay_stream *stream; | |
ad7051c0 | 1641 | struct lttcomm_relayd_quiescent_control msg; |
c8f59ee5 DG |
1642 | struct lttcomm_relayd_generic_reply reply; |
1643 | ||
1644 | DBG("Checking quiescent state on control socket"); | |
1645 | ||
58eb9381 | 1646 | if (!conn->session || conn->version_check_done == 0) { |
ad7051c0 DG |
1647 | ERR("Trying to check for data before version check"); |
1648 | ret = -1; | |
1649 | goto end_no_session; | |
1650 | } | |
1651 | ||
58eb9381 | 1652 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
ad7051c0 | 1653 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1654 | if (ret == 0) { |
1655 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1656 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1657 | } else { |
1658 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1659 | ret); | |
1660 | } | |
ad7051c0 DG |
1661 | ret = -1; |
1662 | goto end_no_session; | |
1663 | } | |
1664 | ||
1665 | stream_id = be64toh(msg.stream_id); | |
5a693d30 MD |
1666 | stream = stream_get_by_id(stream_id); |
1667 | if (!stream) { | |
1668 | goto reply; | |
1669 | } | |
1670 | pthread_mutex_lock(&stream->lock); | |
1671 | stream->data_pending_check_done = true; | |
1672 | pthread_mutex_unlock(&stream->lock); | |
1673 | DBG("Relay quiescent control pending flag set to %" PRIu64, stream_id); | |
1674 | stream_put(stream); | |
1675 | reply: | |
53efb85a | 1676 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1677 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 1678 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1679 | if (ret < 0) { |
6d805429 | 1680 | ERR("Relay data quiescent control ret code failed"); |
c8f59ee5 DG |
1681 | } |
1682 | ||
ad7051c0 | 1683 | end_no_session: |
c8f59ee5 DG |
1684 | return ret; |
1685 | } | |
1686 | ||
f7079f67 | 1687 | /* |
5a693d30 MD |
1688 | * Initialize a data pending command. This means that a consumer is about |
1689 | * to ask for data pending for each stream it holds. Simply iterate over | |
1690 | * all streams of a session and set the data_pending_check_done flag. | |
f7079f67 DG |
1691 | * |
1692 | * This command returns to the client a LTTNG_OK code. | |
1693 | */ | |
5a693d30 | 1694 | static int relay_begin_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1695 | struct relay_connection *conn) |
f7079f67 DG |
1696 | { |
1697 | int ret; | |
1698 | struct lttng_ht_iter iter; | |
1699 | struct lttcomm_relayd_begin_data_pending msg; | |
1700 | struct lttcomm_relayd_generic_reply reply; | |
1701 | struct relay_stream *stream; | |
1702 | uint64_t session_id; | |
1703 | ||
1704 | assert(recv_hdr); | |
58eb9381 | 1705 | assert(conn); |
f7079f67 DG |
1706 | |
1707 | DBG("Init streams for data pending"); | |
1708 | ||
58eb9381 | 1709 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1710 | ERR("Trying to check for data before version check"); |
1711 | ret = -1; | |
1712 | goto end_no_session; | |
1713 | } | |
1714 | ||
58eb9381 | 1715 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1716 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1717 | if (ret == 0) { |
1718 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1719 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1720 | } else { |
1721 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1722 | ret); | |
1723 | } | |
f7079f67 DG |
1724 | ret = -1; |
1725 | goto end_no_session; | |
1726 | } | |
1727 | ||
1728 | session_id = be64toh(msg.session_id); | |
1729 | ||
1730 | /* | |
5a693d30 MD |
1731 | * Iterate over all streams to set the begin data pending flag. |
1732 | * For now, the streams are indexed by stream handle so we have | |
1733 | * to iterate over all streams to find the one associated with | |
1734 | * the right session_id. | |
f7079f67 DG |
1735 | */ |
1736 | rcu_read_lock(); | |
d3e2ba59 | 1737 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1738 | node.node) { |
5a693d30 MD |
1739 | if (!stream_get(stream)) { |
1740 | continue; | |
1741 | } | |
1742 | if (stream->trace->session->id == session_id) { | |
1743 | pthread_mutex_lock(&stream->lock); | |
1744 | stream->data_pending_check_done = false; | |
1745 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 DG |
1746 | DBG("Set begin data pending flag to stream %" PRIu64, |
1747 | stream->stream_handle); | |
1748 | } | |
5a693d30 | 1749 | stream_put(stream); |
f7079f67 DG |
1750 | } |
1751 | rcu_read_unlock(); | |
1752 | ||
53efb85a | 1753 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1754 | /* All good, send back reply. */ |
1755 | reply.ret_code = htobe32(LTTNG_OK); | |
1756 | ||
58eb9381 | 1757 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1758 | if (ret < 0) { |
1759 | ERR("Relay begin data pending send reply failed"); | |
1760 | } | |
1761 | ||
1762 | end_no_session: | |
1763 | return ret; | |
1764 | } | |
1765 | ||
1766 | /* | |
5a693d30 MD |
1767 | * End data pending command. This will check, for a given session id, if |
1768 | * each stream associated with it has its data_pending_check_done flag | |
1769 | * set. If not, this means that the client lost track of the stream but | |
1770 | * the data is still being streamed on our side. In this case, we inform | |
1771 | * the client that data is in flight. | |
f7079f67 DG |
1772 | * |
1773 | * Return to the client if there is data in flight or not with a ret_code. | |
1774 | */ | |
5a693d30 | 1775 | static int relay_end_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1776 | struct relay_connection *conn) |
f7079f67 DG |
1777 | { |
1778 | int ret; | |
1779 | struct lttng_ht_iter iter; | |
1780 | struct lttcomm_relayd_end_data_pending msg; | |
1781 | struct lttcomm_relayd_generic_reply reply; | |
1782 | struct relay_stream *stream; | |
1783 | uint64_t session_id; | |
1784 | uint32_t is_data_inflight = 0; | |
1785 | ||
f7079f67 DG |
1786 | DBG("End data pending command"); |
1787 | ||
58eb9381 | 1788 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1789 | ERR("Trying to check for data before version check"); |
1790 | ret = -1; | |
1791 | goto end_no_session; | |
1792 | } | |
1793 | ||
58eb9381 | 1794 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1795 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1796 | if (ret == 0) { |
1797 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1798 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1799 | } else { |
1800 | ERR("Relay didn't receive valid end data_pending struct size: %d", | |
1801 | ret); | |
1802 | } | |
f7079f67 DG |
1803 | ret = -1; |
1804 | goto end_no_session; | |
1805 | } | |
1806 | ||
1807 | session_id = be64toh(msg.session_id); | |
1808 | ||
5a693d30 MD |
1809 | /* |
1810 | * Iterate over all streams to see if the begin data pending | |
1811 | * flag is set. | |
1812 | */ | |
f7079f67 | 1813 | rcu_read_lock(); |
d3e2ba59 | 1814 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1815 | node.node) { |
5a693d30 MD |
1816 | if (!stream_get(stream)) { |
1817 | continue; | |
1818 | } | |
1819 | if (stream->trace->session->id != session_id) { | |
1820 | stream_put(stream); | |
1821 | continue; | |
1822 | } | |
1823 | pthread_mutex_lock(&stream->lock); | |
1824 | if (!stream->data_pending_check_done) { | |
1825 | if (!stream->closed || !(((int64_t) (stream->prev_seq - stream->last_net_seq_num)) >= 0)) { | |
1826 | is_data_inflight = 1; | |
1827 | DBG("Data is still in flight for stream %" PRIu64, | |
1828 | stream->stream_handle); | |
1829 | pthread_mutex_unlock(&stream->lock); | |
1830 | stream_put(stream); | |
1831 | break; | |
1832 | } | |
f7079f67 | 1833 | } |
5a693d30 MD |
1834 | pthread_mutex_unlock(&stream->lock); |
1835 | stream_put(stream); | |
f7079f67 DG |
1836 | } |
1837 | rcu_read_unlock(); | |
1838 | ||
53efb85a | 1839 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1840 | /* All good, send back reply. */ |
1841 | reply.ret_code = htobe32(is_data_inflight); | |
1842 | ||
58eb9381 | 1843 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1844 | if (ret < 0) { |
1845 | ERR("Relay end data pending send reply failed"); | |
1846 | } | |
1847 | ||
1848 | end_no_session: | |
1849 | return ret; | |
1850 | } | |
1851 | ||
1c20f0e2 JD |
1852 | /* |
1853 | * Receive an index for a specific stream. | |
1854 | * | |
1855 | * Return 0 on success else a negative value. | |
1856 | */ | |
5a693d30 | 1857 | static int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1858 | struct relay_connection *conn) |
1c20f0e2 | 1859 | { |
5a693d30 | 1860 | int ret, send_ret; |
58eb9381 | 1861 | struct relay_session *session = conn->session; |
1c20f0e2 | 1862 | struct lttcomm_relayd_index index_info; |
5a693d30 | 1863 | struct relay_index *index; |
1c20f0e2 JD |
1864 | struct lttcomm_relayd_generic_reply reply; |
1865 | struct relay_stream *stream; | |
1866 | uint64_t net_seq_num; | |
1867 | ||
58eb9381 | 1868 | assert(conn); |
1c20f0e2 JD |
1869 | |
1870 | DBG("Relay receiving index"); | |
1871 | ||
58eb9381 | 1872 | if (!session || conn->version_check_done == 0) { |
1c20f0e2 JD |
1873 | ERR("Trying to close a stream before version check"); |
1874 | ret = -1; | |
1875 | goto end_no_session; | |
1876 | } | |
1877 | ||
58eb9381 | 1878 | ret = conn->sock->ops->recvmsg(conn->sock, &index_info, |
1c20f0e2 JD |
1879 | sizeof(index_info), 0); |
1880 | if (ret < sizeof(index_info)) { | |
1881 | if (ret == 0) { | |
1882 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1883 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
1c20f0e2 JD |
1884 | } else { |
1885 | ERR("Relay didn't receive valid index struct size : %d", ret); | |
1886 | } | |
1887 | ret = -1; | |
1888 | goto end_no_session; | |
1889 | } | |
1890 | ||
1891 | net_seq_num = be64toh(index_info.net_seq_num); | |
1892 | ||
5a693d30 | 1893 | stream = stream_get_by_id(be64toh(index_info.relay_stream_id)); |
1c20f0e2 | 1894 | if (!stream) { |
5a693d30 | 1895 | ERR("stream_get_by_id not found"); |
1c20f0e2 | 1896 | ret = -1; |
5a693d30 | 1897 | goto end; |
1c20f0e2 | 1898 | } |
5a693d30 | 1899 | pthread_mutex_lock(&stream->lock); |
1c20f0e2 | 1900 | |
d3e2ba59 JD |
1901 | /* Live beacon handling */ |
1902 | if (index_info.packet_size == 0) { | |
5a693d30 MD |
1903 | DBG("Received live beacon for stream %" PRIu64, |
1904 | stream->stream_handle); | |
d3e2ba59 JD |
1905 | |
1906 | /* | |
5a693d30 MD |
1907 | * Only flag a stream inactive when it has already |
1908 | * received data and no indexes are in flight. | |
d3e2ba59 | 1909 | */ |
8921c6e4 | 1910 | if (stream->index_received_seqcount > 0 |
5a693d30 MD |
1911 | && stream->indexes_in_flight == 0) { |
1912 | stream->beacon_ts_end = | |
1913 | be64toh(index_info.timestamp_end); | |
d3e2ba59 JD |
1914 | } |
1915 | ret = 0; | |
5a693d30 | 1916 | goto end_stream_put; |
d3e2ba59 JD |
1917 | } else { |
1918 | stream->beacon_ts_end = -1ULL; | |
1919 | } | |
1920 | ||
528f2ffa JD |
1921 | if (stream->ctf_stream_id == -1ULL) { |
1922 | stream->ctf_stream_id = be64toh(index_info.stream_id); | |
1923 | } | |
5a693d30 MD |
1924 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
1925 | if (!index) { | |
1926 | ret = -1; | |
1927 | ERR("relay_index_get_by_id_or_create index NULL"); | |
1928 | goto end_stream_put; | |
1c20f0e2 | 1929 | } |
5a693d30 MD |
1930 | if (set_index_control_data(index, &index_info)) { |
1931 | ERR("set_index_control_data error"); | |
1932 | relay_index_put(index); | |
1933 | ret = -1; | |
1934 | goto end_stream_put; | |
1935 | } | |
1936 | ret = relay_index_try_flush(index); | |
1937 | if (ret == 0) { | |
8921c6e4 MD |
1938 | tracefile_array_commit_seq(stream->tfa); |
1939 | stream->index_received_seqcount++; | |
5a693d30 MD |
1940 | } else if (ret > 0) { |
1941 | /* no flush. */ | |
1942 | ret = 0; | |
1943 | } else { | |
1944 | ERR("relay_index_try_flush error %d", ret); | |
1945 | relay_index_put(index); | |
1946 | ret = -1; | |
1c20f0e2 JD |
1947 | } |
1948 | ||
5a693d30 MD |
1949 | end_stream_put: |
1950 | pthread_mutex_unlock(&stream->lock); | |
1951 | stream_put(stream); | |
1952 | ||
1953 | end: | |
1c20f0e2 | 1954 | |
53efb85a | 1955 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
1956 | if (ret < 0) { |
1957 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1958 | } else { | |
1959 | reply.ret_code = htobe32(LTTNG_OK); | |
1960 | } | |
58eb9381 | 1961 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1c20f0e2 JD |
1962 | if (send_ret < 0) { |
1963 | ERR("Relay sending close index id reply"); | |
1964 | ret = send_ret; | |
1965 | } | |
1966 | ||
1967 | end_no_session: | |
1968 | return ret; | |
1969 | } | |
1970 | ||
a4baae1b JD |
1971 | /* |
1972 | * Receive the streams_sent message. | |
1973 | * | |
1974 | * Return 0 on success else a negative value. | |
1975 | */ | |
5a693d30 | 1976 | static int relay_streams_sent(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1977 | struct relay_connection *conn) |
a4baae1b JD |
1978 | { |
1979 | int ret, send_ret; | |
1980 | struct lttcomm_relayd_generic_reply reply; | |
1981 | ||
58eb9381 | 1982 | assert(conn); |
a4baae1b JD |
1983 | |
1984 | DBG("Relay receiving streams_sent"); | |
1985 | ||
58eb9381 | 1986 | if (!conn->session || conn->version_check_done == 0) { |
a4baae1b JD |
1987 | ERR("Trying to close a stream before version check"); |
1988 | ret = -1; | |
1989 | goto end_no_session; | |
1990 | } | |
1991 | ||
1992 | /* | |
5a693d30 MD |
1993 | * Publish every pending stream in the connection recv list which are |
1994 | * now ready to be used by the viewer. | |
4a9daf17 | 1995 | */ |
5a693d30 | 1996 | publish_connection_local_streams(conn); |
4a9daf17 | 1997 | |
53efb85a | 1998 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 1999 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 2000 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
a4baae1b JD |
2001 | if (send_ret < 0) { |
2002 | ERR("Relay sending sent_stream reply"); | |
2003 | ret = send_ret; | |
2004 | } else { | |
2005 | /* Success. */ | |
2006 | ret = 0; | |
2007 | } | |
2008 | ||
2009 | end_no_session: | |
2010 | return ret; | |
2011 | } | |
2012 | ||
b8aa1682 | 2013 | /* |
d3e2ba59 | 2014 | * Process the commands received on the control socket |
b8aa1682 | 2015 | */ |
5a693d30 | 2016 | static int relay_process_control(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 2017 | struct relay_connection *conn) |
b8aa1682 JD |
2018 | { |
2019 | int ret = 0; | |
2020 | ||
2021 | switch (be32toh(recv_hdr->cmd)) { | |
b8aa1682 | 2022 | case RELAYD_CREATE_SESSION: |
58eb9381 | 2023 | ret = relay_create_session(recv_hdr, conn); |
b8aa1682 | 2024 | break; |
b8aa1682 | 2025 | case RELAYD_ADD_STREAM: |
58eb9381 | 2026 | ret = relay_add_stream(recv_hdr, conn); |
b8aa1682 JD |
2027 | break; |
2028 | case RELAYD_START_DATA: | |
58eb9381 | 2029 | ret = relay_start(recv_hdr, conn); |
b8aa1682 JD |
2030 | break; |
2031 | case RELAYD_SEND_METADATA: | |
58eb9381 | 2032 | ret = relay_recv_metadata(recv_hdr, conn); |
b8aa1682 JD |
2033 | break; |
2034 | case RELAYD_VERSION: | |
58eb9381 | 2035 | ret = relay_send_version(recv_hdr, conn); |
b8aa1682 | 2036 | break; |
173af62f | 2037 | case RELAYD_CLOSE_STREAM: |
58eb9381 | 2038 | ret = relay_close_stream(recv_hdr, conn); |
173af62f | 2039 | break; |
6d805429 | 2040 | case RELAYD_DATA_PENDING: |
58eb9381 | 2041 | ret = relay_data_pending(recv_hdr, conn); |
c8f59ee5 DG |
2042 | break; |
2043 | case RELAYD_QUIESCENT_CONTROL: | |
58eb9381 | 2044 | ret = relay_quiescent_control(recv_hdr, conn); |
c8f59ee5 | 2045 | break; |
f7079f67 | 2046 | case RELAYD_BEGIN_DATA_PENDING: |
58eb9381 | 2047 | ret = relay_begin_data_pending(recv_hdr, conn); |
f7079f67 DG |
2048 | break; |
2049 | case RELAYD_END_DATA_PENDING: | |
58eb9381 | 2050 | ret = relay_end_data_pending(recv_hdr, conn); |
f7079f67 | 2051 | break; |
1c20f0e2 | 2052 | case RELAYD_SEND_INDEX: |
58eb9381 | 2053 | ret = relay_recv_index(recv_hdr, conn); |
1c20f0e2 | 2054 | break; |
a4baae1b | 2055 | case RELAYD_STREAMS_SENT: |
58eb9381 | 2056 | ret = relay_streams_sent(recv_hdr, conn); |
a4baae1b | 2057 | break; |
b8aa1682 JD |
2058 | case RELAYD_UPDATE_SYNC_INFO: |
2059 | default: | |
2060 | ERR("Received unknown command (%u)", be32toh(recv_hdr->cmd)); | |
58eb9381 | 2061 | relay_unknown_command(conn); |
b8aa1682 JD |
2062 | ret = -1; |
2063 | goto end; | |
2064 | } | |
2065 | ||
2066 | end: | |
2067 | return ret; | |
2068 | } | |
2069 | ||
7d2f7452 DG |
2070 | /* |
2071 | * Handle index for a data stream. | |
2072 | * | |
5a693d30 | 2073 | * Called with the stream lock held. |
7d2f7452 DG |
2074 | * |
2075 | * Return 0 on success else a negative value. | |
2076 | */ | |
2077 | static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, | |
2078 | int rotate_index) | |
2079 | { | |
5a693d30 MD |
2080 | int ret = 0; |
2081 | uint64_t data_offset; | |
2082 | struct relay_index *index; | |
7d2f7452 | 2083 | |
7d2f7452 DG |
2084 | /* Get data offset because we are about to update the index. */ |
2085 | data_offset = htobe64(stream->tracefile_size_current); | |
2086 | ||
de948e0e MD |
2087 | DBG("handle_index_data: stream %" PRIu64 " net_seq_num %" PRIu64 " data offset %" PRIu64, |
2088 | stream->stream_handle, net_seq_num, stream->tracefile_size_current); | |
5a693d30 | 2089 | |
7d2f7452 | 2090 | /* |
5a693d30 MD |
2091 | * Lookup for an existing index for that stream id/sequence |
2092 | * number. If it exists, the control thread has already received the | |
2093 | * data for it, thus we need to write it to disk. | |
7d2f7452 | 2094 | */ |
5a693d30 | 2095 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
7d2f7452 | 2096 | if (!index) { |
5a693d30 MD |
2097 | ret = -1; |
2098 | goto end; | |
7d2f7452 DG |
2099 | } |
2100 | ||
5a693d30 MD |
2101 | if (rotate_index || !stream->index_fd) { |
2102 | int fd; | |
2103 | ||
2104 | /* Put ref on previous index_fd. */ | |
2105 | if (stream->index_fd) { | |
2106 | stream_fd_put(stream->index_fd); | |
2107 | stream->index_fd = NULL; | |
7d2f7452 | 2108 | } |
7d2f7452 | 2109 | |
5a693d30 MD |
2110 | fd = index_create_file(stream->path_name, stream->channel_name, |
2111 | -1, -1, stream->tracefile_size, | |
8921c6e4 | 2112 | tracefile_array_get_file_index_head(stream->tfa)); |
5a693d30 MD |
2113 | if (fd < 0) { |
2114 | ret = -1; | |
2115 | /* Put self-ref for this index due to error. */ | |
2116 | relay_index_put(index); | |
2117 | goto end; | |
2118 | } | |
2119 | stream->index_fd = stream_fd_create(fd); | |
2120 | if (!stream->index_fd) { | |
2121 | ret = -1; | |
2122 | if (close(fd)) { | |
2123 | PERROR("Error closing FD %d", fd); | |
2124 | } | |
2125 | /* Put self-ref for this index due to error. */ | |
2126 | relay_index_put(index); | |
2127 | /* Will put the local ref. */ | |
2128 | goto end; | |
7d2f7452 | 2129 | } |
7d2f7452 DG |
2130 | } |
2131 | ||
5a693d30 MD |
2132 | if (relay_index_set_fd(index, stream->index_fd, data_offset)) { |
2133 | ret = -1; | |
2134 | /* Put self-ref for this index due to error. */ | |
2135 | relay_index_put(index); | |
2136 | goto end; | |
7d2f7452 DG |
2137 | } |
2138 | ||
5a693d30 MD |
2139 | ret = relay_index_try_flush(index); |
2140 | if (ret == 0) { | |
8921c6e4 MD |
2141 | tracefile_array_commit_seq(stream->tfa); |
2142 | stream->index_received_seqcount++; | |
5a693d30 MD |
2143 | } else if (ret > 0) { |
2144 | /* No flush. */ | |
2145 | ret = 0; | |
2146 | } else { | |
2147 | /* Put self-ref for this index due to error. */ | |
2148 | relay_index_put(index); | |
2149 | ret = -1; | |
2150 | } | |
2151 | end: | |
7d2f7452 DG |
2152 | return ret; |
2153 | } | |
2154 | ||
b8aa1682 JD |
2155 | /* |
2156 | * relay_process_data: Process the data received on the data socket | |
2157 | */ | |
5a693d30 | 2158 | static int relay_process_data(struct relay_connection *conn) |
b8aa1682 | 2159 | { |
7d2f7452 | 2160 | int ret = 0, rotate_index = 0; |
6cd525e8 | 2161 | ssize_t size_ret; |
b8aa1682 JD |
2162 | struct relay_stream *stream; |
2163 | struct lttcomm_relayd_data_hdr data_hdr; | |
7d2f7452 | 2164 | uint64_t stream_id; |
173af62f | 2165 | uint64_t net_seq_num; |
b8aa1682 | 2166 | uint32_t data_size; |
2a174661 | 2167 | struct relay_session *session; |
a08107cc | 2168 | bool new_stream = false, close_requested = false; |
b8aa1682 | 2169 | |
58eb9381 | 2170 | ret = conn->sock->ops->recvmsg(conn->sock, &data_hdr, |
7c5aef62 | 2171 | sizeof(struct lttcomm_relayd_data_hdr), 0); |
b8aa1682 | 2172 | if (ret <= 0) { |
a6cd2b97 DG |
2173 | if (ret == 0) { |
2174 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2175 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2176 | } else { |
58eb9381 | 2177 | ERR("Unable to receive data header on sock %d", conn->sock->fd); |
a6cd2b97 | 2178 | } |
b8aa1682 JD |
2179 | ret = -1; |
2180 | goto end; | |
2181 | } | |
2182 | ||
2183 | stream_id = be64toh(data_hdr.stream_id); | |
5a693d30 | 2184 | stream = stream_get_by_id(stream_id); |
b8aa1682 | 2185 | if (!stream) { |
de948e0e | 2186 | ERR("relay_process_data: Cannot find stream %" PRIu64, stream_id); |
b8aa1682 | 2187 | ret = -1; |
5a693d30 | 2188 | goto end; |
b8aa1682 | 2189 | } |
5a693d30 | 2190 | session = stream->trace->session; |
b8aa1682 JD |
2191 | data_size = be32toh(data_hdr.data_size); |
2192 | if (data_buffer_size < data_size) { | |
c617c0c6 MD |
2193 | char *tmp_data_ptr; |
2194 | ||
2195 | tmp_data_ptr = realloc(data_buffer, data_size); | |
2196 | if (!tmp_data_ptr) { | |
b8aa1682 | 2197 | ERR("Allocating data buffer"); |
c617c0c6 | 2198 | free(data_buffer); |
b8aa1682 | 2199 | ret = -1; |
5a693d30 | 2200 | goto end_stream_put; |
b8aa1682 | 2201 | } |
c617c0c6 | 2202 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
2203 | data_buffer_size = data_size; |
2204 | } | |
2205 | memset(data_buffer, 0, data_size); | |
2206 | ||
173af62f DG |
2207 | net_seq_num = be64toh(data_hdr.net_seq_num); |
2208 | ||
77c7c900 | 2209 | DBG3("Receiving data of size %u for stream id %" PRIu64 " seqnum %" PRIu64, |
173af62f | 2210 | data_size, stream_id, net_seq_num); |
58eb9381 | 2211 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
b8aa1682 | 2212 | if (ret <= 0) { |
a6cd2b97 DG |
2213 | if (ret == 0) { |
2214 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2215 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
de948e0e MD |
2216 | } else { |
2217 | ERR("Socket %d error %d", conn->sock->fd, ret); | |
a6cd2b97 | 2218 | } |
b8aa1682 | 2219 | ret = -1; |
5a693d30 | 2220 | goto end_stream_put; |
b8aa1682 JD |
2221 | } |
2222 | ||
5a693d30 MD |
2223 | pthread_mutex_lock(&stream->lock); |
2224 | ||
1c20f0e2 | 2225 | /* Check if a rotation is needed. */ |
0f907de1 JD |
2226 | if (stream->tracefile_size > 0 && |
2227 | (stream->tracefile_size_current + data_size) > | |
2228 | stream->tracefile_size) { | |
8921c6e4 MD |
2229 | uint64_t old_id, new_id; |
2230 | ||
2231 | old_id = tracefile_array_get_file_index_head(stream->tfa); | |
2232 | tracefile_array_file_rotate(stream->tfa); | |
2233 | ||
2234 | /* new_id is updated by utils_rotate_stream_file. */ | |
2235 | new_id = old_id; | |
6b6b9a5a | 2236 | |
5a693d30 MD |
2237 | ret = utils_rotate_stream_file(stream->path_name, |
2238 | stream->channel_name, stream->tracefile_size, | |
2239 | stream->tracefile_count, -1, | |
2240 | -1, stream->stream_fd->fd, | |
8921c6e4 | 2241 | &new_id, &stream->stream_fd->fd); |
0f907de1 | 2242 | if (ret < 0) { |
1c20f0e2 | 2243 | ERR("Rotating stream output file"); |
5a693d30 MD |
2244 | goto end_stream_unlock; |
2245 | } | |
5a693d30 MD |
2246 | /* |
2247 | * Reset current size because we just performed a stream | |
2248 | * rotation. | |
2249 | */ | |
a6976990 | 2250 | stream->tracefile_size_current = 0; |
1c20f0e2 JD |
2251 | rotate_index = 1; |
2252 | } | |
2253 | ||
1c20f0e2 | 2254 | /* |
5a693d30 MD |
2255 | * Index are handled in protocol version 2.4 and above. Also, |
2256 | * snapshot and index are NOT supported. | |
1c20f0e2 | 2257 | */ |
2a174661 | 2258 | if (session->minor >= 4 && !session->snapshot) { |
7d2f7452 | 2259 | ret = handle_index_data(stream, net_seq_num, rotate_index); |
1c20f0e2 | 2260 | if (ret < 0) { |
de948e0e MD |
2261 | ERR("handle_index_data: fail stream %" PRIu64 " net_seq_num %" PRIu64 " ret %d", |
2262 | stream->stream_handle, net_seq_num, ret); | |
5a693d30 | 2263 | goto end_stream_unlock; |
1c20f0e2 | 2264 | } |
1c20f0e2 JD |
2265 | } |
2266 | ||
7d2f7452 | 2267 | /* Write data to stream output fd. */ |
5a693d30 | 2268 | size_ret = lttng_write(stream->stream_fd->fd, data_buffer, data_size); |
6cd525e8 | 2269 | if (size_ret < data_size) { |
b8aa1682 JD |
2270 | ERR("Relay error writing data to file"); |
2271 | ret = -1; | |
5a693d30 | 2272 | goto end_stream_unlock; |
b8aa1682 | 2273 | } |
1d4dfdef | 2274 | |
5a693d30 MD |
2275 | DBG2("Relay wrote %zd bytes to tracefile for stream id %" PRIu64, |
2276 | size_ret, stream->stream_handle); | |
5ab7344e | 2277 | |
5a693d30 MD |
2278 | ret = write_padding_to_file(stream->stream_fd->fd, |
2279 | be32toh(data_hdr.padding_size)); | |
1d4dfdef | 2280 | if (ret < 0) { |
de948e0e MD |
2281 | ERR("write_padding_to_file: fail stream %" PRIu64 " net_seq_num %" PRIu64 " ret %d", |
2282 | stream->stream_handle, net_seq_num, ret); | |
5a693d30 | 2283 | goto end_stream_unlock; |
1d4dfdef | 2284 | } |
5a693d30 MD |
2285 | stream->tracefile_size_current += |
2286 | data_size + be32toh(data_hdr.padding_size); | |
1b95be80 MD |
2287 | if (stream->prev_seq == -1ULL) { |
2288 | new_stream = true; | |
2289 | } | |
2290 | ||
173af62f DG |
2291 | stream->prev_seq = net_seq_num; |
2292 | ||
5a693d30 | 2293 | end_stream_unlock: |
a08107cc | 2294 | close_requested = stream->close_requested; |
5a693d30 | 2295 | pthread_mutex_unlock(&stream->lock); |
a08107cc JG |
2296 | if (close_requested) { |
2297 | try_stream_close(stream); | |
2298 | } | |
2299 | ||
1b95be80 MD |
2300 | if (new_stream) { |
2301 | pthread_mutex_lock(&session->lock); | |
2302 | uatomic_set(&session->new_streams, 1); | |
2303 | pthread_mutex_unlock(&session->lock); | |
2304 | } | |
5a693d30 MD |
2305 | end_stream_put: |
2306 | stream_put(stream); | |
b8aa1682 JD |
2307 | end: |
2308 | return ret; | |
2309 | } | |
2310 | ||
5a693d30 | 2311 | static void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
2312 | { |
2313 | int ret; | |
2314 | ||
58eb9381 | 2315 | (void) lttng_poll_del(events, pollfd); |
b8aa1682 JD |
2316 | |
2317 | ret = close(pollfd); | |
2318 | if (ret < 0) { | |
2319 | ERR("Closing pollfd %d", pollfd); | |
2320 | } | |
2321 | } | |
2322 | ||
5a693d30 MD |
2323 | static void relay_thread_close_connection(struct lttng_poll_event *events, |
2324 | int pollfd, struct relay_connection *conn) | |
9d1bbf21 | 2325 | { |
5a693d30 | 2326 | const char *type_str; |
2a174661 | 2327 | |
5a693d30 MD |
2328 | switch (conn->type) { |
2329 | case RELAY_DATA: | |
2330 | type_str = "Data"; | |
2331 | break; | |
2332 | case RELAY_CONTROL: | |
2333 | type_str = "Control"; | |
2334 | break; | |
2335 | case RELAY_VIEWER_COMMAND: | |
2336 | type_str = "Viewer Command"; | |
2337 | break; | |
2338 | case RELAY_VIEWER_NOTIFICATION: | |
2339 | type_str = "Viewer Notification"; | |
2340 | break; | |
2341 | default: | |
2342 | type_str = "Unknown"; | |
9d1bbf21 | 2343 | } |
5a693d30 MD |
2344 | cleanup_connection_pollfd(events, pollfd); |
2345 | connection_put(conn); | |
2346 | DBG("%s connection closed with %d", type_str, pollfd); | |
b8aa1682 JD |
2347 | } |
2348 | ||
2349 | /* | |
2350 | * This thread does the actual work | |
2351 | */ | |
5a693d30 | 2352 | static void *relay_thread_worker(void *data) |
b8aa1682 | 2353 | { |
beaad64c DG |
2354 | int ret, err = -1, last_seen_data_fd = -1; |
2355 | uint32_t nb_fd; | |
b8aa1682 JD |
2356 | struct lttng_poll_event events; |
2357 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 | 2358 | struct lttng_ht_iter iter; |
b8aa1682 | 2359 | struct lttcomm_relayd_hdr recv_hdr; |
90e7d72f | 2360 | struct relay_connection *destroy_conn = NULL; |
b8aa1682 JD |
2361 | |
2362 | DBG("[thread] Relay worker started"); | |
2363 | ||
9d1bbf21 MD |
2364 | rcu_register_thread(); |
2365 | ||
55706a7d MD |
2366 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
2367 | ||
9b5e0863 MD |
2368 | if (testpoint(relayd_thread_worker)) { |
2369 | goto error_testpoint; | |
2370 | } | |
2371 | ||
f385ae0a MD |
2372 | health_code_update(); |
2373 | ||
b8aa1682 JD |
2374 | /* table of connections indexed on socket */ |
2375 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
2376 | if (!relay_connections_ht) { |
2377 | goto relay_connections_ht_error; | |
2378 | } | |
b8aa1682 | 2379 | |
b8aa1682 JD |
2380 | ret = create_thread_poll_set(&events, 2); |
2381 | if (ret < 0) { | |
2382 | goto error_poll_create; | |
2383 | } | |
2384 | ||
58eb9381 | 2385 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
2386 | if (ret < 0) { |
2387 | goto error; | |
2388 | } | |
2389 | ||
beaad64c | 2390 | restart: |
b8aa1682 | 2391 | while (1) { |
beaad64c DG |
2392 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
2393 | ||
f385ae0a MD |
2394 | health_code_update(); |
2395 | ||
b8aa1682 | 2396 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 2397 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 2398 | health_poll_entry(); |
b8aa1682 | 2399 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 2400 | health_poll_exit(); |
b8aa1682 JD |
2401 | if (ret < 0) { |
2402 | /* | |
2403 | * Restart interrupted system call. | |
2404 | */ | |
2405 | if (errno == EINTR) { | |
2406 | goto restart; | |
2407 | } | |
2408 | goto error; | |
2409 | } | |
2410 | ||
0d9c5d77 DG |
2411 | nb_fd = ret; |
2412 | ||
beaad64c | 2413 | /* |
5a693d30 MD |
2414 | * Process control. The control connection is |
2415 | * prioritized so we don't starve it with high | |
2416 | * throughput tracing data on the data connection. | |
beaad64c | 2417 | */ |
b8aa1682 JD |
2418 | for (i = 0; i < nb_fd; i++) { |
2419 | /* Fetch once the poll data */ | |
beaad64c DG |
2420 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
2421 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 2422 | |
f385ae0a MD |
2423 | health_code_update(); |
2424 | ||
fd20dac9 | 2425 | if (!revents) { |
5a693d30 MD |
2426 | /* |
2427 | * No activity for this FD (poll | |
2428 | * implementation). | |
2429 | */ | |
fd20dac9 MD |
2430 | continue; |
2431 | } | |
2432 | ||
b8aa1682 JD |
2433 | /* Thread quit pipe has been closed. Killing thread. */ |
2434 | ret = check_thread_quit_pipe(pollfd, revents); | |
2435 | if (ret) { | |
095a4ae5 MD |
2436 | err = 0; |
2437 | goto exit; | |
b8aa1682 JD |
2438 | } |
2439 | ||
58eb9381 DG |
2440 | /* Inspect the relay conn pipe for new connection */ |
2441 | if (pollfd == relay_conn_pipe[0]) { | |
92d6debb | 2442 | if (revents & LPOLLIN) { |
90e7d72f JG |
2443 | struct relay_connection *conn; |
2444 | ||
58eb9381 | 2445 | ret = lttng_read(relay_conn_pipe[0], &conn, sizeof(conn)); |
b8aa1682 JD |
2446 | if (ret < 0) { |
2447 | goto error; | |
2448 | } | |
58eb9381 DG |
2449 | lttng_poll_add(&events, conn->sock->fd, |
2450 | LPOLLIN | LPOLLRDHUP); | |
5a693d30 | 2451 | connection_ht_add(relay_connections_ht, conn); |
58eb9381 | 2452 | DBG("Connection socket %d added", conn->sock->fd); |
92d6debb MD |
2453 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
2454 | ERR("Relay connection pipe error"); | |
2455 | goto error; | |
2456 | } else { | |
2457 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
2458 | goto error; | |
b8aa1682 | 2459 | } |
58eb9381 | 2460 | } else { |
90e7d72f JG |
2461 | struct relay_connection *ctrl_conn; |
2462 | ||
5a693d30 | 2463 | ctrl_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
58eb9381 | 2464 | /* If not found, there is a synchronization issue. */ |
90e7d72f | 2465 | assert(ctrl_conn); |
58eb9381 | 2466 | |
92d6debb MD |
2467 | if (ctrl_conn->type == RELAY_DATA) { |
2468 | if (revents & LPOLLIN) { | |
beaad64c DG |
2469 | /* |
2470 | * Flag the last seen data fd not deleted. It will be | |
2471 | * used as the last seen fd if any fd gets deleted in | |
2472 | * this first loop. | |
2473 | */ | |
2474 | last_notdel_data_fd = pollfd; | |
2475 | } | |
92d6debb MD |
2476 | goto put_ctrl_connection; |
2477 | } | |
2478 | assert(ctrl_conn->type == RELAY_CONTROL); | |
2479 | ||
2480 | if (revents & LPOLLIN) { | |
2481 | ret = ctrl_conn->sock->ops->recvmsg(ctrl_conn->sock, | |
2482 | &recv_hdr, sizeof(recv_hdr), 0); | |
2483 | if (ret <= 0) { | |
2484 | /* Connection closed */ | |
2485 | relay_thread_close_connection(&events, pollfd, | |
2486 | ctrl_conn); | |
2487 | } else { | |
2488 | ret = relay_process_control(&recv_hdr, ctrl_conn); | |
2489 | if (ret < 0) { | |
2490 | /* Clear the session on error. */ | |
2491 | relay_thread_close_connection(&events, | |
2492 | pollfd, ctrl_conn); | |
2493 | } | |
2494 | seen_control = 1; | |
2495 | } | |
2496 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
2497 | relay_thread_close_connection(&events, | |
2498 | pollfd, ctrl_conn); | |
2499 | if (last_seen_data_fd == pollfd) { | |
2500 | last_seen_data_fd = last_notdel_data_fd; | |
2501 | } | |
58eb9381 | 2502 | } else { |
92d6debb MD |
2503 | ERR("Unexpected poll events %u for control sock %d", |
2504 | revents, pollfd); | |
2505 | connection_put(ctrl_conn); | |
2506 | goto error; | |
beaad64c | 2507 | } |
92d6debb | 2508 | put_ctrl_connection: |
5a693d30 | 2509 | connection_put(ctrl_conn); |
beaad64c DG |
2510 | } |
2511 | } | |
2512 | ||
2513 | /* | |
2514 | * The last loop handled a control request, go back to poll to make | |
2515 | * sure we prioritise the control socket. | |
2516 | */ | |
2517 | if (seen_control) { | |
2518 | continue; | |
2519 | } | |
2520 | ||
2521 | if (last_seen_data_fd >= 0) { | |
2522 | for (i = 0; i < nb_fd; i++) { | |
2523 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
2524 | |
2525 | health_code_update(); | |
2526 | ||
beaad64c DG |
2527 | if (last_seen_data_fd == pollfd) { |
2528 | idx = i; | |
2529 | break; | |
2530 | } | |
2531 | } | |
2532 | } | |
2533 | ||
2534 | /* Process data connection. */ | |
2535 | for (i = idx + 1; i < nb_fd; i++) { | |
2536 | /* Fetch the poll data. */ | |
2537 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
2538 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
90e7d72f | 2539 | struct relay_connection *data_conn; |
beaad64c | 2540 | |
f385ae0a MD |
2541 | health_code_update(); |
2542 | ||
fd20dac9 MD |
2543 | if (!revents) { |
2544 | /* No activity for this FD (poll implementation). */ | |
2545 | continue; | |
2546 | } | |
2547 | ||
beaad64c | 2548 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 2549 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
2550 | continue; |
2551 | } | |
2552 | ||
5a693d30 | 2553 | data_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
90e7d72f | 2554 | if (!data_conn) { |
fd20dac9 | 2555 | /* Skip it. Might be removed before. */ |
fd20dac9 MD |
2556 | continue; |
2557 | } | |
92d6debb MD |
2558 | if (data_conn->type == RELAY_CONTROL) { |
2559 | goto put_data_connection; | |
2560 | } | |
2561 | assert(data_conn->type == RELAY_DATA); | |
fd20dac9 MD |
2562 | |
2563 | if (revents & LPOLLIN) { | |
90e7d72f | 2564 | ret = relay_process_data(data_conn); |
fd20dac9 MD |
2565 | /* Connection closed */ |
2566 | if (ret < 0) { | |
5a693d30 | 2567 | relay_thread_close_connection(&events, pollfd, |
92d6debb | 2568 | data_conn); |
fd20dac9 MD |
2569 | /* |
2570 | * Every goto restart call sets the last seen fd where | |
2571 | * here we don't really care since we gracefully | |
2572 | * continue the loop after the connection is deleted. | |
2573 | */ | |
2574 | } else { | |
2575 | /* Keep last seen port. */ | |
2576 | last_seen_data_fd = pollfd; | |
5a693d30 | 2577 | connection_put(data_conn); |
fd20dac9 | 2578 | goto restart; |
b8aa1682 | 2579 | } |
92d6debb MD |
2580 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
2581 | relay_thread_close_connection(&events, pollfd, | |
2582 | data_conn); | |
2583 | } else { | |
2584 | ERR("Unknown poll events %u for data sock %d", | |
2585 | revents, pollfd); | |
b8aa1682 | 2586 | } |
92d6debb | 2587 | put_data_connection: |
5a693d30 | 2588 | connection_put(data_conn); |
b8aa1682 | 2589 | } |
beaad64c | 2590 | last_seen_data_fd = -1; |
b8aa1682 JD |
2591 | } |
2592 | ||
f385ae0a MD |
2593 | /* Normal exit, no error */ |
2594 | ret = 0; | |
2595 | ||
095a4ae5 | 2596 | exit: |
b8aa1682 | 2597 | error: |
58eb9381 | 2598 | /* Cleanup reamaining connection object. */ |
9d1bbf21 | 2599 | rcu_read_lock(); |
90e7d72f JG |
2600 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, |
2601 | destroy_conn, | |
58eb9381 | 2602 | sock_n.node) { |
f385ae0a | 2603 | health_code_update(); |
5a693d30 MD |
2604 | /* |
2605 | * No need to grab another ref, because we own | |
2606 | * destroy_conn. | |
2607 | */ | |
2608 | relay_thread_close_connection(&events, destroy_conn->sock->fd, | |
2609 | destroy_conn); | |
b8aa1682 | 2610 | } |
94d49140 | 2611 | rcu_read_unlock(); |
5a693d30 MD |
2612 | |
2613 | lttng_poll_clean(&events); | |
7d2f7452 | 2614 | error_poll_create: |
b8aa1682 | 2615 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 2616 | relay_connections_ht_error: |
58eb9381 DG |
2617 | /* Close relay conn pipes */ |
2618 | utils_close_pipe(relay_conn_pipe); | |
095a4ae5 MD |
2619 | if (err) { |
2620 | DBG("Thread exited with error"); | |
2621 | } | |
b8aa1682 | 2622 | DBG("Worker thread cleanup complete"); |
095a4ae5 | 2623 | free(data_buffer); |
9b5e0863 | 2624 | error_testpoint: |
f385ae0a MD |
2625 | if (err) { |
2626 | health_error(); | |
2627 | ERR("Health error occurred in %s", __func__); | |
2628 | } | |
2629 | health_unregister(health_relayd); | |
9d1bbf21 | 2630 | rcu_unregister_thread(); |
b4aacfdc | 2631 | lttng_relay_stop_threads(); |
b8aa1682 JD |
2632 | return NULL; |
2633 | } | |
2634 | ||
2635 | /* | |
2636 | * Create the relay command pipe to wake thread_manage_apps. | |
2637 | * Closed in cleanup(). | |
2638 | */ | |
58eb9381 | 2639 | static int create_relay_conn_pipe(void) |
b8aa1682 | 2640 | { |
a02de639 | 2641 | int ret; |
b8aa1682 | 2642 | |
58eb9381 | 2643 | ret = utils_create_pipe_cloexec(relay_conn_pipe); |
b8aa1682 | 2644 | |
b8aa1682 JD |
2645 | return ret; |
2646 | } | |
2647 | ||
2648 | /* | |
2649 | * main | |
2650 | */ | |
2651 | int main(int argc, char **argv) | |
2652 | { | |
178a0557 | 2653 | int ret = 0, retval = 0; |
b8aa1682 JD |
2654 | void *status; |
2655 | ||
b8aa1682 JD |
2656 | /* Parse arguments */ |
2657 | progname = argv[0]; | |
178a0557 MD |
2658 | if (set_options(argc, argv)) { |
2659 | retval = -1; | |
2660 | goto exit_options; | |
b8aa1682 JD |
2661 | } |
2662 | ||
178a0557 MD |
2663 | if (set_signal_handler()) { |
2664 | retval = -1; | |
2665 | goto exit_options; | |
b8aa1682 JD |
2666 | } |
2667 | ||
4d513a50 DG |
2668 | /* Try to create directory if -o, --output is specified. */ |
2669 | if (opt_output_path) { | |
994fa64f DG |
2670 | if (*opt_output_path != '/') { |
2671 | ERR("Please specify an absolute path for -o, --output PATH"); | |
178a0557 MD |
2672 | retval = -1; |
2673 | goto exit_options; | |
994fa64f DG |
2674 | } |
2675 | ||
d85144f4 JG |
2676 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG, |
2677 | -1, -1); | |
4d513a50 DG |
2678 | if (ret < 0) { |
2679 | ERR("Unable to create %s", opt_output_path); | |
178a0557 MD |
2680 | retval = -1; |
2681 | goto exit_options; | |
4d513a50 DG |
2682 | } |
2683 | } | |
2684 | ||
b8aa1682 | 2685 | /* Daemonize */ |
b5218ffb | 2686 | if (opt_daemon || opt_background) { |
3fd27398 MD |
2687 | int i; |
2688 | ||
2689 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, | |
2690 | !opt_background); | |
b8aa1682 | 2691 | if (ret < 0) { |
178a0557 MD |
2692 | retval = -1; |
2693 | goto exit_options; | |
b8aa1682 | 2694 | } |
3fd27398 MD |
2695 | |
2696 | /* | |
2697 | * We are in the child. Make sure all other file | |
2698 | * descriptors are closed, in case we are called with | |
2699 | * more opened file descriptors than the standard ones. | |
2700 | */ | |
2701 | for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { | |
2702 | (void) close(i); | |
2703 | } | |
2704 | } | |
2705 | ||
178a0557 MD |
2706 | |
2707 | /* Initialize thread health monitoring */ | |
2708 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
2709 | if (!health_relayd) { | |
2710 | PERROR("health_app_create error"); | |
2711 | retval = -1; | |
2712 | goto exit_health_app_create; | |
2713 | } | |
2714 | ||
3fd27398 | 2715 | /* Create thread quit pipe */ |
178a0557 MD |
2716 | if (init_thread_quit_pipe()) { |
2717 | retval = -1; | |
2718 | goto exit_init_data; | |
b8aa1682 JD |
2719 | } |
2720 | ||
1c20f0e2 | 2721 | /* Check if daemon is UID = 0 */ |
5a693d30 | 2722 | if (!getuid()) { |
8d5c808e | 2723 | if (control_uri->port < 1024 || data_uri->port < 1024 || live_uri->port < 1024) { |
b8aa1682 | 2724 | ERR("Need to be root to use ports < 1024"); |
178a0557 MD |
2725 | retval = -1; |
2726 | goto exit_init_data; | |
b8aa1682 JD |
2727 | } |
2728 | } | |
2729 | ||
2730 | /* Setup the thread apps communication pipe. */ | |
178a0557 MD |
2731 | if (create_relay_conn_pipe()) { |
2732 | retval = -1; | |
2733 | goto exit_init_data; | |
b8aa1682 JD |
2734 | } |
2735 | ||
2736 | /* Init relay command queue. */ | |
8bdee6e2 | 2737 | cds_wfcq_init(&relay_conn_queue.head, &relay_conn_queue.tail); |
b8aa1682 JD |
2738 | |
2739 | /* Set up max poll set size */ | |
25b397f9 MD |
2740 | if (lttng_poll_set_max_size()) { |
2741 | retval = -1; | |
2742 | goto exit_init_data; | |
2743 | } | |
b8aa1682 | 2744 | |
554831e7 MD |
2745 | /* Initialize communication library */ |
2746 | lttcomm_init(); | |
87e45c13 | 2747 | lttcomm_inet_init(); |
554831e7 | 2748 | |
d3e2ba59 | 2749 | /* tables of sessions indexed by session ID */ |
5a693d30 MD |
2750 | sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2751 | if (!sessions_ht) { | |
178a0557 MD |
2752 | retval = -1; |
2753 | goto exit_init_data; | |
d3e2ba59 JD |
2754 | } |
2755 | ||
2756 | /* tables of streams indexed by stream ID */ | |
2a174661 | 2757 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 | 2758 | if (!relay_streams_ht) { |
178a0557 MD |
2759 | retval = -1; |
2760 | goto exit_init_data; | |
d3e2ba59 JD |
2761 | } |
2762 | ||
2763 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
2764 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2765 | if (!viewer_streams_ht) { | |
178a0557 MD |
2766 | retval = -1; |
2767 | goto exit_init_data; | |
55706a7d MD |
2768 | } |
2769 | ||
65931c8b | 2770 | ret = utils_create_pipe(health_quit_pipe); |
178a0557 MD |
2771 | if (ret) { |
2772 | retval = -1; | |
2773 | goto exit_health_quit_pipe; | |
65931c8b MD |
2774 | } |
2775 | ||
2776 | /* Create thread to manage the client socket */ | |
2777 | ret = pthread_create(&health_thread, NULL, | |
2778 | thread_manage_health, (void *) NULL); | |
178a0557 MD |
2779 | if (ret) { |
2780 | errno = ret; | |
65931c8b | 2781 | PERROR("pthread_create health"); |
178a0557 MD |
2782 | retval = -1; |
2783 | goto exit_health_thread; | |
65931c8b MD |
2784 | } |
2785 | ||
b8aa1682 JD |
2786 | /* Setup the dispatcher thread */ |
2787 | ret = pthread_create(&dispatcher_thread, NULL, | |
2788 | relay_thread_dispatcher, (void *) NULL); | |
178a0557 MD |
2789 | if (ret) { |
2790 | errno = ret; | |
b8aa1682 | 2791 | PERROR("pthread_create dispatcher"); |
178a0557 MD |
2792 | retval = -1; |
2793 | goto exit_dispatcher_thread; | |
b8aa1682 JD |
2794 | } |
2795 | ||
2796 | /* Setup the worker thread */ | |
2797 | ret = pthread_create(&worker_thread, NULL, | |
5a693d30 | 2798 | relay_thread_worker, NULL); |
178a0557 MD |
2799 | if (ret) { |
2800 | errno = ret; | |
b8aa1682 | 2801 | PERROR("pthread_create worker"); |
178a0557 MD |
2802 | retval = -1; |
2803 | goto exit_worker_thread; | |
b8aa1682 JD |
2804 | } |
2805 | ||
2806 | /* Setup the listener thread */ | |
2807 | ret = pthread_create(&listener_thread, NULL, | |
2808 | relay_thread_listener, (void *) NULL); | |
178a0557 MD |
2809 | if (ret) { |
2810 | errno = ret; | |
b8aa1682 | 2811 | PERROR("pthread_create listener"); |
178a0557 MD |
2812 | retval = -1; |
2813 | goto exit_listener_thread; | |
b8aa1682 JD |
2814 | } |
2815 | ||
5a693d30 | 2816 | ret = relayd_live_create(live_uri); |
178a0557 | 2817 | if (ret) { |
d3e2ba59 | 2818 | ERR("Starting live viewer threads"); |
178a0557 | 2819 | retval = -1; |
50138f51 | 2820 | goto exit_live; |
d3e2ba59 JD |
2821 | } |
2822 | ||
178a0557 MD |
2823 | /* |
2824 | * This is where we start awaiting program completion (e.g. through | |
2825 | * signal that asks threads to teardown). | |
2826 | */ | |
2827 | ||
2828 | ret = relayd_live_join(); | |
2829 | if (ret) { | |
2830 | retval = -1; | |
2831 | } | |
50138f51 | 2832 | exit_live: |
178a0557 | 2833 | |
b8aa1682 | 2834 | ret = pthread_join(listener_thread, &status); |
178a0557 MD |
2835 | if (ret) { |
2836 | errno = ret; | |
2837 | PERROR("pthread_join listener_thread"); | |
2838 | retval = -1; | |
b8aa1682 JD |
2839 | } |
2840 | ||
178a0557 | 2841 | exit_listener_thread: |
b8aa1682 | 2842 | ret = pthread_join(worker_thread, &status); |
178a0557 MD |
2843 | if (ret) { |
2844 | errno = ret; | |
2845 | PERROR("pthread_join worker_thread"); | |
2846 | retval = -1; | |
b8aa1682 JD |
2847 | } |
2848 | ||
178a0557 | 2849 | exit_worker_thread: |
b8aa1682 | 2850 | ret = pthread_join(dispatcher_thread, &status); |
178a0557 MD |
2851 | if (ret) { |
2852 | errno = ret; | |
2853 | PERROR("pthread_join dispatcher_thread"); | |
2854 | retval = -1; | |
b8aa1682 | 2855 | } |
178a0557 | 2856 | exit_dispatcher_thread: |
42415026 | 2857 | |
65931c8b | 2858 | ret = pthread_join(health_thread, &status); |
178a0557 MD |
2859 | if (ret) { |
2860 | errno = ret; | |
2861 | PERROR("pthread_join health_thread"); | |
2862 | retval = -1; | |
65931c8b | 2863 | } |
178a0557 | 2864 | exit_health_thread: |
65931c8b | 2865 | |
65931c8b | 2866 | utils_close_pipe(health_quit_pipe); |
178a0557 | 2867 | exit_health_quit_pipe: |
65931c8b | 2868 | |
178a0557 | 2869 | exit_init_data: |
55706a7d | 2870 | health_app_destroy(health_relayd); |
55706a7d | 2871 | exit_health_app_create: |
178a0557 | 2872 | exit_options: |
5a693d30 MD |
2873 | relayd_cleanup(); |
2874 | ||
2875 | /* Ensure all prior call_rcu are done. */ | |
2876 | rcu_barrier(); | |
d3e2ba59 | 2877 | |
178a0557 | 2878 | if (!retval) { |
b8aa1682 | 2879 | exit(EXIT_SUCCESS); |
178a0557 MD |
2880 | } else { |
2881 | exit(EXIT_FAILURE); | |
b8aa1682 | 2882 | } |
b8aa1682 | 2883 | } |